From mailer@bialko.llnl.gov Mon May 1 17:17:04 2000 Return-Path: Date: Mon, 1 May 2000 17:15:07 -0700 (PDT) From: mailer@bialko.llnl.gov (Automatic Reply) To: karplus@cse.ucsc.edu Subject: CASP4 Predictor Registration CASP4: Predictor Registration ============================= Thank you for registering your interest in the Fourth Critical Assessment of Techniques for Protein Structure Prediction Experiment (CASP4). For your records, a summary of your registration is printed below. Please save and print out this file for your records. Keep a note of your Predictor Registration Code. You will need this to update your registration details and/or submit predictions. Predictor Registration Code = 3670-4530-6947 1. Name Kevin Karplus 2. Name of Institution University of California, Santa Cruz 3. Mailing address Street: Computer Engineering City: Santa Cruz State: CA Zip: 95064 Country: USA 4. Telephone 1-831-459-4250 5. Fax 1-831-459-4829 6. Email karplus@cse.ucsc.edu 7. Registration type group leader 8. Group Name (e.g. Smith-John) SAM_T2K 9. Likelyhood of submitting predictions under different categories of the experiment Comparative Modelling: Likely Fold Recognition ('threading'): Certain Ab initio: Certain 10. Source of information about experiment participated in CASP2 and CASP3 From compbio-request Mon May 1 17:20:14 2000 Return-Path: Date: Mon, 1 May 2000 17:19:48 -0700 From: Kevin Karplus To: compbio@cse.ucsc.edu Subject: CASP4 predictor registration I have registered as a group leader for CASP4 as group SAM_T2K Anyone who has provided substantial input to the method, or who plans to help out with the structure prediction this summer, should let me know and register as a group member using web site http://PredictionCenter.llnl.gov/casp4/predictors/forms/casp4-reg.html If you need the registration code, it is in pc/experiments/casp4/mail From dfischer@cs.bgu.ac.il Tue May 2 06:31:53 2000 Return-Path: From: Daniel Fischer Date: Tue, 2 May 2000 16:31:41 +0300 (IDT) To: karplus@cse.ucsc.edu Dear potential CAFASP-2 participant, The CAFASP-2 experiment is about to start. This means that very soon the CAFASP meta-server will begin submitting the targets. In order to participate in CAFASP-2 you must FIRST register your server at the CASP-4 Prediction Center web site at http://PredictionCenter.llnl.gov/casp4/ (use the special form for CAFASP), and SUBSEQUENTLY, register your server at the CAFASP-2 meta-server site at http://cafasp.bioinfo.pl/. Upon release of each target, the CAFASP meta-server will submit it to only those servers that have registered at both sites. Servers that register late, will automatically get a "no prediction" for all previous targets. Soon, we will release the third CAFASP-2 announcement describing in more detail the CAFASP procedures. So, if you intend to participate in CAFASP, you are still on time to register without loosing targets. From karplus@cse.ucsc.edu Tue May 2 09:27:08 2000 Return-Path: Date: Tue, 2 May 2000 09:27:00 -0700 From: Kevin Karplus To: dfischer@cs.bgu.ac.il CC: karplus@cse.ucsc.edu In-reply-to: <200005021331.QAA02604@lace.cs.bgu.ac.il> (message from Daniel Fischer on Tue, 2 May 2000 16:31:41 +0300 (IDT)) Subject: CAFASP2 registration I had previously registered SAM_T99 (the T99-query.html website) with CAFASP, and have now entered a SAM_T99 group on the CASP4/CAFASP2 web form. Do I need to do anything further? From rph@cse.ucsc.edu Tue May 2 09:31:29 2000 Return-Path: From: "rph" To: "Kevin Karplus" Subject: FW: CASP4 Predictor Registration Date: Tue, 2 May 2000 09:31:28 -0700 Content-Type: text/plain; charset="iso-8859-1" X-Priority: 3 (Normal) X-MSMail-Priority: Normal Importance: Normal X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2919.6700 -----Original Message----- From: Automatic Reply [mailto:mailer@bialko.llnl.gov] Sent: Tuesday, May 02, 2000 9:26 AM To: sam-info@cse.ucsc.edu Subject: CASP4 Predictor Registration CASP4: Predictor Registration ============================= Thank you for registering your interest in the Fourth Critical Assessment of Techniques for Protein Structure Prediction Experiment (CASP4). For your records, a summary of your registration is printed below. Please save and print out this file for your records. Keep a note of your Predictor Registration Code. You will need this to update your registration details and/or submit predictions. Predictor Registration Code = 2000-8506-6614 1. Name Kevin Karplus , http://www.cse.ucsc.edu/research/compbio/T99-query.html 2. Name of Institution University of California, Santa Cruz 3. Mailing address Street: Computer Engineering City: Santa Cruz State: CA Zip: 95064 Country: USA 4. Telephone 1-831-459-4250 5. Fax 1-831-459-4829 6. Email sam-info@cse.ucsc.edu 7. Registration type group leader 8. Group Name (e.g. Smith-John) SAM_T99 9. Likelyhood of submitting predictions under different categories of the experiment Comparative Modelling: Certain Fold Recognition ('threading'): Certain Ab initio: Certain 10. Source of information about experiment participated in CAFASP1 From rph@cse.ucsc.edu Tue May 2 09:33:30 2000 Return-Path: From: "rph" To: "Kevin Karplus" Subject: FW: CASP4 Predictor Registration Date: Tue, 2 May 2000 09:33:29 -0700 Content-Type: text/plain; charset="iso-8859-1" X-Priority: 3 (Normal) X-MSMail-Priority: Normal Importance: Normal X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2919.6700 -----Original Message----- From: Automatic Reply [mailto:mailer@bialko.llnl.gov] Sent: Tuesday, May 02, 2000 9:24 AM To: sam-info@cse.ucsc.edu Subject: CASP4 Predictor Registration CASP4: Predictor Registration ============================= Thank you for registering your interest in the Fourth Critical Assessment of Techniques for Protein Structure Prediction Experiment (CASP4). For your records, a summary of your registration is printed below. Please save and print out this file for your records. Keep a note of your Predictor Registration Code. You will need this to update your registration details and/or submit predictions. Predictor Registration Code = 3074-2981-6622 1. Name Kevin Karplus , http://www.cse.ucsc.edu/research/compbio/T99-query.html 2. Name of Institution University of California, Santa Cruz 3. Mailing address Street: Computer Engineering City: Santa Cruz State: CA Zip: 95064 Country: USA 4. Telephone 1-831-459-4250 5. Fax 1-831-459-4829 6. Email sam-info@cse.ucsc.edu 7. Registration type group leader 8. Group Name (e.g. Smith-John) SAM_T99 9. Likelyhood of submitting predictions under different categories of the experiment Comparative Modelling: Certain Fold Recognition ('threading'): Certain Ab initio: Certain 10. Source of information about experiment participated in CAFASP1 From dfischer@cs.bgu.ac.il Wed May 3 09:51:46 2000 Return-Path: X-Authentication-Warning: lead.cs.bgu.ac.il: dfischer owned process doing -bs Date: Wed, 3 May 2000 19:51:19 +0300 (IDT) From: Daniel Fischer X-Sender: dfischer@lead To: Kevin Karplus Subject: Re: CAFASP2 registration In-Reply-To: <200005021627.JAA25652@purr.cse.ucsc.edu> Content-Type: TEXT/PLAIN; charset=US-ASCII Kevin, I didn't know about your bike accident until I got your automated reply. I wish you a quick full recovery! > I had previously registered SAM_T99 (the T99-query.html website) with > CAFASP, and have now entered a SAM_T99 group on the CASP4/CAFASP2 web form. > Do I need to do anything further? Just send the group id and name you got from the prediction center to Leszek. Are you entering only one of your methods? I thought you had 2 or 3. ------------------ Dr. Daniel Fischer Dept. of Math and Computer Science Ben Gurion University, Beer-Sheva 84015, Israel Tel +972-7-6472714 e-mail: dfischer@cs.bgu.ac.il Fax +972-7-6477648 or 6472910 http://www.cs.bgu.ac.il/~dfischer From karplus@cse.ucsc.edu Wed May 3 12:33:41 2000 Return-Path: Date: Wed, 3 May 2000 12:33:24 -0700 From: Kevin Karplus To: dfischer@cs.bgu.ac.il CC: karplus@cse.ucsc.edu In-reply-to: (message from Daniel Fischer on Wed, 3 May 2000 19:51:19 +0300 (IDT)) Subject: Re: CAFASP2 registration Thanks for the kind thoughts---I am recovering well from the bike accident. The ribs have mostly healed, and I'm just waiting for the incision in the abdomen to finish healing. I hope to be back on my bike in about a month. We'll be entering only the combined method, for simplicity. I considered submitting each of the methods independently, but two of them are not good enough in our tests to be worth the trouble. I'm not sure whether the combined method is better or worse than the best of the methods in the combination, but I'm not interested enough to submit them separately (the difference will be fairly small, I think). I hope that we'll have a new method implemented in the next couple of weeks, that will not be part of CAFASP, but will be part of CASP. From karplus@cse.ucsc.edu Wed May 3 12:35:00 2000 Return-Path: Date: Wed, 3 May 2000 12:34:55 -0700 From: Kevin Karplus To: leszek@bioinfo.pl CC: karplus@cse.ucsc.edu Subject: CAFASP registration I have registered our server with CASP4 as SAM_T99. You already have the information about how to run the server (I assume it is all working for you now). Do you need any further information from me? From dfischer@cs.bgu.ac.il Wed May 3 13:25:01 2000 Return-Path: X-Authentication-Warning: lace.cs.bgu.ac.il: dfischer owned process doing -bs Date: Wed, 3 May 2000 23:24:30 +0300 (IDT) From: Daniel Fischer X-Sender: dfischer@lace To: Kevin Karplus Subject: Re: CAFASP2 registration In-Reply-To: <200005031933.MAA28208@purr.cse.ucsc.edu> Content-Type: TEXT/PLAIN; charset=US-ASCII > Thanks for the kind thoughts---I am recovering well from the bike accident. > The ribs have mostly healed, and I'm just waiting for the incision in > the abdomen to finish healing.I hope to be back on my bike in about > a month. Again, I wish you a quick recovery! > > I hope that we'll have a new method implemented in the next couple of > weeks, that will not be part of CAFASP, but will be part of CASP. Whenever your new method is ready, you can enter it into cafasp, even if it may miss some targets, only to get an additional evaluation. If you won't have it ready within a server, notice that you can still enter it into the CAFASP within the automated-programs-without-server type of submissions, again for evaluation purposes only. Our announcement describes this category. Dani ------------------ Dr. Daniel Fischer Dept. of Math and Computer Science Ben Gurion University, Beer-Sheva 84015, Israel Tel +972-7-6472714 e-mail: dfischer@cs.bgu.ac.il Fax +972-7-6477648 or 6472910 http://www.cs.bgu.ac.il/~dfischer From leszek@bioinfo.pl Wed May 3 14:56:32 2000 Return-Path: Date: Wed, 03 May 2000 23:54:13 +0200 From: Leszek Rychlewski X-Accept-Language: en To: Kevin Karplus Subject: Re: CAFASP registration Content-Type: text/plain; charset=us-ascii Kevin Karplus wrote: > I have registered our server with CASP4 as SAM_T99. > You already have the information about how to run the server (I assume > it is all working for you now). > > Do you need any further information from me? Everything works perfect. I will let You know if any problems occur. Best regards Leszek PS: You are of course registered. From leszek@bioinfo.pl Fri May 19 02:56:07 2000 Return-Path: Sender: leszek@cyborg.bioinfo.pl Date: Fri, 19 May 2000 11:52:06 +0200 From: Leszek Rychlewski X-Accept-Language: en To: Kevin Karplus CC: stu@cse.ucsc.edu Subject: Re: I don't see T0089 Content-Type: text/plain; charset=iso-8859-15 Kevin Karplus wrote: > The T0089 results just appeared on http://cafasp.bioinfo.pl/target/ > > We had some problems with our file server this week, and our web site > may have had jobs killed. > > If our jobs did not go through the first time, please resubmit them. Dear Kevin, I just resubmitted the first target. I will submit the next one after I get the first result. Yours, Leszek PS: if there are other major problems You can always run the prediction Yourself and send me the results via E-mail (promising that they were the result of the automated procedure). PPS: Logs: Thu May 18 21:59:04 CEST 2000 job farmer.958680174.b at Thu May 18 13: 02:54 2000 Connection: close Content-Type: text/html Submission Confirmation

Submission Confirmation

Your request has been received.

The results will be sent to T0089.target99@cafasp.bioinfo.pl

Fri May 19 11:43:28 CEST 2000 job farmer.958729638.b at Fri May 19 02: 47:18 2000 Connection: close Content-Type: text/html Submission Confirmation

Submission Confirmation

Your request has been received.

The results will be sent to T0089.target99@cafasp.bioinfo.pl

From karplus@cse.ucsc.edu Fri May 19 10:20:48 2000 Return-Path: Date: Fri, 19 May 2000 10:20:44 -0700 From: Kevin Karplus To: leszek@bioinfo.pl CC: stu@cse.ucsc.edu, karplus@cse.ucsc.edu In-reply-to: <39250EC6.767D9B6B@bioinfo.pl> (message from Leszek Rychlewski on Fri, 19 May 2000 11:52:06 +0200) Subject: Re: I don't see T0089 The problem was that the human genome annotation turned on all 100 machines of the new cluster, and ran jobs that made so many read an write requests to the same file server that it crashed. Unfortunately, that is the same file server that supports our web site, so the jobs running at the time also died. The problem occurred several times before they figured out that the file server wasn't having hardware problems---they were just asking far too much from it. They finally figured out how to put the temporary files from RepeatMasker on the local disks, rather than the file server, and things are back to normal (at least for a while). It is unfortunate that the outage occurred during the narrow window for some CAFASP predictions, but as long as you can re-run missing results, we'll take the "late" mark. In general, our predictor should never refuse to give any result---if you don't get a result in 24 hours, there is probably something wrong, and the job should be resubmitted. Kevin From leszek@bioinfo.pl Fri May 19 10:35:53 2000 Return-Path: Sender: leszek@cyborg.bioinfo.pl Date: Fri, 19 May 2000 19:31:58 +0200 From: Leszek Rychlewski X-Accept-Language: en To: Kevin Karplus CC: stu@cse.ucsc.edu Subject: Re: I don't see T0089 Content-Type: text/plain; charset=iso-8859-15 Kevin Karplus wrote: > The problem was that the human genome annotation turned on all 100 > machines of the new cluster, and ran jobs that made so many read an > write requests to the same file server that it crashed. > Unfortunately, that is the same file server that supports our web > site, so the jobs running at the time also died. > > The problem occurred several times before they figured out that the > file server wasn't having hardware problems---they were just asking > far too much from it. They finally figured out how to put the > temporary files from RepeatMasker on the local disks, rather than the > file server, and things are back to normal (at least for a while). > > It is unfortunate that the outage occurred during the narrow window > for some CAFASP predictions, but as long as you can re-run missing > results, we'll take the "late" mark. > > In general, our predictor should never refuse to give any result---if > you don't get a result in 24 hours, there is probably something wrong, > and the job should be resubmitted. > > Kevin I just submitted T0090 . You are not "late" until tomorrow evening. From our side the resubmission of jobs is no problem. Please let me know if the new job is running. Yours, Leszek From leszek@bioinfo.pl Fri May 19 10:37:33 2000 Return-Path: Sender: leszek@cyborg.bioinfo.pl Date: Fri, 19 May 2000 19:33:26 +0200 From: Leszek Rychlewski X-Accept-Language: en To: Kevin Karplus CC: stu@cse.ucsc.edu Subject: Re: I don't see T0089 Content-Type: text/plain; charset=us-ascii T0090 is running, I already got first results. I will submit T0089 shortly. Leszek From karplus@cse.ucsc.edu Fri May 19 13:08:35 2000 Return-Path: Date: Fri, 19 May 2000 13:08:34 -0700 From: Kevin Karplus To: stu@cse.ucsc.edu CC: leszek@bioinfo.pl, karplus@cse.ucsc.edu Subject: serious bug in web server I think everyone is getting back a blank "hits" file, based on the info in pc/tmp/target99-query/output. This is not right, as I know that T0090 is easily close enough to get 1MUT and 1TUM as high-scoring hits. The secondary structure files don't appear to have been returned since mid-day yesterday either. Spencer, please check to see what is wrong with your web script--we'll probably have to take a "late submission" hit on these sequences, since the web script seems to have broken. I could not find any files that had changed since May 17, so I don't know what changed to break things. When you have found and fixed the bug, please send mail to leszek@bioinfo.pl, so that the targets can be resubmitted (albeit late). Also send messages to everyone else who has used the server since 11am May 18, since all their runs may need to be redone (only the t99 multiple alignment seems to be returned correctly right now). From leszek@bioinfo.pl Fri May 19 08:31:12 2000 Return-Path: Sender: leszek@cyborg.bioinfo.pl Date: Fri, 19 May 2000 17:27:17 +0200 From: Leszek Rychlewski X-Accept-Language: en To: Kevin Karplus CC: stu@cse.ucsc.edu Subject: Re: I don't see T0089 Content-Type: text/plain; charset=us-ascii Kevin Karplus wrote: > The T0089 results just appeared on http://cafasp.bioinfo.pl/target/ > > We had some problems with our file server this week, and our web site > may have had jobs killed. > > If our jobs did not go through the first time, please resubmit them. Dear Kevin, I still did not obtain any results. I will submit the next targets in few hours. I will submit both again tomorrow morning if I did not get any results. Please send the result via E-mail to me if we can not find the problem. Leszek From dfischer@cs.bgu.ac.il Tue May 23 10:11:49 2000 Return-Path: X-Authentication-Warning: lace.cs.bgu.ac.il: dfischer owned process doing -bs Date: Tue, 23 May 2000 20:10:52 +0300 (IDT) From: Daniel Fischer X-Sender: dfischer@lace To: leszek@bioinfo.pl, kelley@icrf.icnet.uk, David.Jones@brunel.ac.uk, karplus@cse.ucsc.edu, kenji@cryst.bioc.cam.ac.uk, wjb@dnamining.com, arussell@rsc.anu.edu.au Subject: cafasp summaries Content-Type: MULTIPART/MIXED; BOUNDARY="-559023410-851401618-959101852=:27634" This message is in MIME format. The first part should be readable text, while the remaining parts are likely unreadable without MIME-aware tools. Send mail to mime@docserver.cac.washington.edu for more info. ---559023410-851401618-959101852=:27634 Content-Type: TEXT/PLAIN; charset=US-ASCII Dear CAFASPers, We are creating for each of the fold recognition targets a summary web-page that allows easy visualization of each of the results stored at the CAFASP meta-server. These summary pages will be made publicly available from the CAFASP site. In these summary pages we will list the top 5 hits of each server along with the reported scores. In addition, to ease the analysis of the results, we annotate the top hits with the corresponding SCOP ids. To this purpose we would like you to confirm that we have correctly interpreted the raw-output of your server. 1. In the table below, would you acknowledge that the scoring and significance threshold listed for your server is correct? 2. In the html file attached, there is a sample summary page for target T0086. Would you please check that the top hits and scores listed do correspond to your server's output? For some of the servers it is obvious, but for others, some parsing is required, and we would like to be sure that this is correct before we proceed with the other targets. Thanks for your cooperation. If we do not receive a reply from you within a couple of days, we will assume that the information is ok. THRESHOLDS PDB-BLAST Significance Cutoff < 0.1 FFAS Contact E-mail : adam@burnham-inst.org Primary Score Parameter : Z-score Significance Cutoff > 8.0 3D-PSSM Contact E-mail : kelley@icrf.icnet.uk Primary Score Parameter : E-value Significance Cutoff < 0.37 GENTHREADER and mGENTHREADER Contact E-mail : David.Jones@brunel.ac.uk Primary Score Paramater : unknown Significance Cutoff : unknown BIOINBGU Contact E-mail : dfischer@cs.bgu.ac.il Primary Score Parameter : Z-score Significance cutoff : * gonp > 6 * gonpm > 6 * seqpprf > 7 * seqpmprf > 7 * pfseq > 6 * consensus : o 19+ most probably certain o 15-19 highly probable o 12-15 probable only if at least one of the five methods is above the corresponding threshold SAM T-98 and SAM T-99 Contact E-mail : karplus@cse.ucsc.edu Primary Score Parameter : E-value Significance Cutoff < approx 0.01 FUGUE Contact E-Mail : kenji@cryst.bioc.cam.ac.uk Primary Score Parameters : Z-score Significance Cutoff > 6 P - MAP Contact E-mail : wjb@dnamining.com Primary Score Parameters : E-value Significance Cutoff < 0.01 SAUSAGE Contact E-mail : arussell@rsc.anu.edu.au Primary Score Parameters : Rank Significance Cutoff : UNKNOWN ------------------ Dr. Daniel Fischer Dept. of Math and Computer Science Ben Gurion University, Beer-Sheva 84015, Israel Tel +972-7-6472714 e-mail: dfischer@cs.bgu.ac.il Fax +972-7-6477648 or 6472910 http://www.cs.bgu.ac.il/~dfischer ---559023410-851401618-959101852=:27634 Content-Type: TEXT/PLAIN; charset=US-ASCII; name="T0086.html" Content-ID: Content-Description: T0086 summary in html Content-Disposition: attachment; filename="T0086.html" Content-Transfer-Encoding: BASE64 PGh0bWw+DQogICAgIDxoZWFkPg0KICAgICAgICAgIDx0aXRsZT50MDA4NiB0 YWJsZTwvdGl0bGU+DQogICAgIDwvaGVhZD4NCg0KPGJvZHkgYmdjb2xvcj0j NzM5YWM0ID4NCiAgICAgPGgyPjxzdHJvbmc+PGNlbnRlciA+VDAwODY8L2Nl bnRlcj48L3N0cm9uZz48L2gyPg0KICAgICA8dGFibGUgYm9yZGVyPSIxIiBj ZWxscGFkZGluZz0iNSIgY2Vsc3BhY2luZz0iNSIgd2lkdGg9IjYwJSI+DQog ICAgICAgICAgPHRyPg0KICAgICAgICAgICAgIDx0aD4gc2VydmVyIG5hbWUN CiAgICAgICAgICAgICA8dGg+IGhpdDENCiAgICAgICAgICAgICA8dGg+IHNj b3JlMQ0KICAgICAgICAgICAgIDx0aD4gaGl0Mg0KICAgICAgICAgICAgIDx0 aD4gc2NvcmUyDQogICAgICAgICAgICAgPHRoPiBoaXQzDQogICAgICAgICAg ICAgPHRoPiBzY29yZTMNCiAgICAgICAgICAgICA8dGg+IGhpdDQNCiAgICAg ICAgICAgICA8dGg+IHNjb3JlNA0KICAgICAgICAgICAgIDx0aD4gaGl0NQ0K ICAgICAgICAgICAgIDx0aD4gc2NvcmU1DQogICAgICAgICAgICAgPHRoPiBM aXN0IG9mIHRvcCA1IHBkYiBjb2Rlcw0KICAgICAgICAgIDx0cj4NCg0KDQog ICAgICAgICAgICAgPHRkPlBEQi1CTEFTVA0KICAgICAgICAgICAgIDx0ZD48 c3Ryb25nPjMuNDcuNQ0KICAgICAgICAgICAgIDx0ZD4xLjgNCiAgICAgICAg ICAgICA8dGQ+PHN0cm9uZz4zLjQ3LjUNCiAgICAgICAgICAgICA8dGQ+MS44 DQogICAgICAgICAgICAgPHRkPjxzdHJvbmc+MWI2ZmENCiAgICAgICAgICAg ICA8dGQ+Mi4zDQogICAgICAgICAgICAgPHRkPjxzdHJvbmc+MmNhbUINCiAg ICAgICAgICAgICA8dGQ+Mi4zDQogICAgICAgICAgICAgPHRkPjxzdHJvbmc+ MWJ3ekENCiAgICAgICAgICAgICA8dGQ+NS4yDQogICAgICAgICAgICAgPFRE PiAxYm91QiAxYjR1QiAxZHR5QiAxZHR5QSAxZDVhQQ0KICAgICAgICAgIDx0 cj4NCg0KDQogICAgICAgICAgICAgPHRkPkZTU0ENCiAgICAgICAgICAgICA8 dGQ+PHN0cm9uZz4xLjg1LjENCiAgICAgICAgICAgICA8dGQ+NC41OQ0KICAg ICAgICAgICAgIDx0ZD48c3Ryb25nPjFhdjRfMw0KICAgICAgICAgICAgIDx0 ZD40LjUxDQogICAgICAgICAgICAgPHRkPjxzdHJvbmc+MXd0dUENCiAgICAg ICAgICAgICA8dGQ+My45Ng0KICAgICAgICAgICAgIDx0ZD48c3Ryb25nPjIu NTQuMQ0KICAgICAgICAgICAgIDx0ZD4zLjk2DQogICAgICAgICAgICAgPHRk PjxzdHJvbmc+NC4xOS4xDQogICAgICAgICAgICAgPHRkPjMuNzYNCiAgICAg ICAgICAgICA8VEQ+IDFib3VBIDFxamdGIDFiNnRBIDJjYW1CIDFid3pBDQog ICAgICAgICAgPHRyPg0KDQoNCiAgICAgICAgICAgICA8dGQ+M0RfUFNTTQ0K ICAgICAgICAgICAgIDx0ZD48c3Ryb25nPjQuNDcuNw0KICAgICAgICAgICAg IDx0ZD4yLjk2DQogICAgICAgICAgICAgPHRkPjxzdHJvbmc+NC4xNS4yDQog ICAgICAgICAgICAgPHRkPjQuMDcNCiAgICAgICAgICAgICA8dGQ+PHN0cm9u Zz4xLjU3LjENCiAgICAgICAgICAgICA8dGQ+NC4zNw0KICAgICAgICAgICAg IDx0ZD48c3Ryb25nPjMuMzQuMQ0KICAgICAgICAgICAgIDx0ZD40Ljg2DQog ICAgICAgICAgICAgPHRkPjxzdHJvbmc+NC40Ny43DQogICAgICAgICAgICAg PHRkPjUuMjANCiAgICAgICAgICAgICA8VEQ+IGMzcHJvY18gZDFoYTFfMSBk MWF2NF8zIGQxd3R1YV8gZDFjenBhXw0KICAgICAgICAgIDx0cj4NCg0KDQog ICAgICAgICAgICAgPHRkPkdFTlRIUkVBREVSDQogICAgICAgICAgICAgPHRk PjxzdHJvbmc+MS4xMDIuMQ0KICAgICAgICAgICAgIDx0ZD4wLjE1OQ0KICAg ICAgICAgICAgIDx0ZD48c3Ryb25nPjMuMzguMQ0KICAgICAgICAgICAgIDx0 ZD4wLjA2Ng0KICAgICAgICAgICAgIDx0ZD48c3Ryb25nPjMuMjQuMQ0KICAg ICAgICAgICAgIDx0ZD4wLjA1Nw0KICAgICAgICAgICAgIDx0ZD48c3Ryb25n PjQuNzEuMQ0KICAgICAgICAgICAgIDx0ZD4wLjA1Mw0KICAgICAgICAgICAg IDx0ZD48c3Ryb25nPjIuNDEuMQ0KICAgICAgICAgICAgIDx0ZD4wLjAzOA0K ICAgICAgICAgICAgIDxURD4gMWpoZ0EwIDFnbHFBMiAxZ3NhMDEgMWM5MUEx IDFhMWVEMg0KICAgICAgICAgIDx0cj4NCg0KDQogICAgICAgICAgICAgPHRk PiBtLUdFTlRIUkVBREVSDQogICAgICAgICAgICAgPHRkPiA8c3Ryb25nPjEu MTAyLjENCiAgICAgICAgICAgICA8dGQ+IDAuMTU5DQogICAgICAgICAgICAg PHRkPiA8c3Ryb25nPjQuMTMzLjENCiAgICAgICAgICAgICA8dGQ+IDAuMDY2 DQogICAgICAgICAgICAgPHRkPiA8c3Ryb25nPjEuMi40DQogICAgICAgICAg ICAgPHRkPiAwLjA2Mg0KICAgICAgICAgICAgIDx0ZD4gPHN0cm9uZz4xLjQ4 LjENCiAgICAgICAgICAgICA8dGQ+IDAuMDU3DQogICAgICAgICAgICAgPHRk PiA8c3Ryb25nPjMuMzguMQ0KICAgICAgICAgICAgIDx0ZD4gMC4wNTUNCiAg ICAgICAgICAgICA8VEQ+IDFqaGdBMCAxYTBwMDIgMXNldEExIDFnc2EwMSAx Z2xxQTINCiAgICAgICAgICA8dHI+IA0KDQoNCiAgICAgICAgICAgICA8dGQ+ YmlvaW5iZ3UoZ29ucCkNCiAgICAgICAgICAgICA8dGQ+PHN0cm9uZz4xcWFy DQogICAgICAgICAgICAgPHRkPjQuNA0KICAgICAgICAgICAgIDx0ZD48c3Ry b25nPjMuODAuMQ0KICAgICAgICAgICAgIDx0ZD4zLjYNCiAgICAgICAgICAg ICA8dGQ+PHN0cm9uZz40Ljc0LjINCiAgICAgICAgICAgICA8dGQ+My40DQog ICAgICAgICAgICAgPHRkPjxzdHJvbmc+Mi4zNi4xDQogICAgICAgICAgICAg ICAgIDxzdHJvbmc+Ni4yLjENCiAgICAgICAgICAgICA8dGQ+My40DQogICAg ICAgICAgICAgPHRkPjxzdHJvbmc+NC4yLjENCiAgICAgICAgICAgICA8dGQ+ My4yDQogICAgICAgICAgICAgPFREPiAxcWFyIDFjYmYgMWFmNSAxYWk5IDFs ejENCiAgICAgICAgICA8dHI+ICAgDQoNCg0KICAgICAgICAgICAgIDx0ZD5i aW9pbmJndShnb25wbSkNCiAgICAgICAgICAgICA8dGQ+PHN0cm9uZz40LjQ3 LjYNCiAgICAgICAgICAgICA8dGQ+My45DQogICAgICAgICAgICAgPHRkPjxz dHJvbmc+NC41Mi4yDQogICAgICAgICAgICAgPHRkPjMuOA0KICAgICAgICAg ICAgIDx0ZD48c3Ryb25nPjQuMi4xDQogICAgICAgICAgICAgPHRkPjMuNg0K ICAgICAgICAgICAgIDx0ZD48c3Ryb25nPjQuNzQuMg0KICAgICAgICAgICAg IDx0ZD4zLjQNCiAgICAgICAgICAgICA8dGQ+PHN0cm9uZz4xcWFyDQogICAg ICAgICAgICAgPHRkPjMuNA0KICAgICAgICAgICAgIDxURD4gMW5kayAxZXBz IDFsejEgMWFmNSAxcWFyDQogICAgICAgICAgPHRyPiAgDQoNCg0KICAgICAg ICAgICAgIDx0ZD5iaW9pbmJndShzZXFwcHJmKQ0KICAgICAgICAgICAgIDx0 ZD48c3Ryb25nPjUuMTcuMQ0KICAgICAgICAgICAgIDx0ZD40LjUNCiAgICAg ICAgICAgICA8dGQ+PHN0cm9uZz4xLjEwMi4xDQogICAgICAgICAgICAgPHRk PjQuMA0KICAgICAgICAgICAgIDx0ZD48c3Ryb25nPjMuMy4xDQogICAgICAg ICAgICAgPHRkPjMuNw0KICAgICAgICAgICAgIDx0ZD48c3Ryb25nPjFjbXoN CiAgICAgICAgICAgICA8dGQ+My43DQogICAgICAgICAgICAgPHRkPjxzdHJv bmc+MXFtcA0KICAgICAgICAgICAgIDx0ZD4zLjYNCiAgICAgICAgICAgICA8 VEQ+IDFqdWQgMndycCAzZ3JzIDFjbXogMXFtcA0KICAgICAgICAgIDx0cj4g ICANCg0KDQogICAgICAgICAgICAgPHRkPmJpb2luYmd1KHNlcXBtcHJmKQ0K ICAgICAgICAgICAgIDxUSD4gTk8NCiAgICAgICAgICAgICA8VEg+IFJFU1VM VA0KICAgICAgICAgICAgIDxUSD4gUkVDRUlWRUQNCiAgICAgICAgICA8dHI+ ICAgDQoNCg0KICAgICAgICAgICAgIDx0ZD4gYmlvaW5iZ3UoY29uc2Vuc3Vz KQ0KICAgICAgICAgICAgIDx0ZD4gPHN0cm9uZz40Ljc0LjINCiAgICAgICAg ICAgICA8dGQ+IDcuNQ0KICAgICAgICAgICAgIDx0ZD4gPHN0cm9uZz4xcXJh Yw0KICAgICAgICAgICAgIDx0ZD4gNi40DQogICAgICAgICAgICAgPHRkPiA8 c3Ryb25nPjQuNDcuNg0KICAgICAgICAgICAgIDx0ZD4gNS42DQogICAgICAg ICAgICAgPHRkPiA8c3Ryb25nPjUuMTcuMQ0KICAgICAgICAgICAgIDx0ZD4g NS42DQogICAgICAgICAgICAgPHRkPiA8c3Ryb25nPjQuNTIuMg0KICAgICAg ICAgICAgIDx0ZD4gMy43DQogICAgICAgICAgICAgPFREPiAxYWY1IDFxYXJj IDFuZGsgMWp1ZCAxZXBzMg0KICAgICAgICAgIDx0cj4NCg0KDQogICAgICAg ICAgICAgPHRkPlNBTS1UOTgNCiAgICAgICAgICAgICA8dGQ+PHN0cm9uZz4x LjQuMTxicj4NCiAgICAgICAgICAgICAgICAgPHN0cm9uZz4xLjM2LjENCiAg ICAgICAgICAgICA8dGQ+IC03LjcxDQogICAgICAgICAgICAgPHRkPiA8c3Ry b25nPjMuMzYuMg0KICAgICAgICAgICAgIDx0ZD4gLTcuNDcNCiAgICAgICAg ICAgICA8dGQ+IDxzdHJvbmc+Mi4zMC42DQogICAgICAgICAgICAgPHRkPiAt Ny4zOA0KICAgICAgICAgICAgIDx0ZD4gPHN0cm9uZz42LjEuMQ0KICAgICAg ICAgICAgIDx0ZD4gLTcuMTUNCiAgICAgICAgICAgICA8dGQ+IDxzdHJvbmc+ MS4yNy4xDQogICAgICAgICAgICAgPHRkPiAtNy4wOA0KICAgICAgICAgICAg IDxURD4gMWF1N0EgMWlpYkEgMWFoakIgMWNvbEEgMmlsaw0KICAgICAgICAg IDx0cj4NCg0KDQogICAgICAgICAgICAgPHRkPiBTQU0tVDk5DQogICAgICAg ICAgICAgPHRkPiA8c3Ryb25nPjEuMi40PGJyPg0KICAgICAgICAgICAgICAg ICAgPHN0cm9uZz40LjgyLjENCiAgICAgICAgICAgICA8dGQ+IDAuNzU4MyAg ICANCiAgICAgICAgICAgICA8dGQ+IDxzdHJvbmc+MS4xMDAuMQ0KICAgICAg ICAgICAgIDx0ZD4gNC41NDENCiAgICAgICAgICAgICA8dGQ+IDxzdHJvbmc+ NC4xMDYuMQ0KICAgICAgICAgICAgIDx0ZD4gNS4xMg0KICAgICAgICAgICAg IDx0ZD4gPHN0cm9uZz40LjEzMC4xDQogICAgICAgICAgICAgPHRkPiA2Ljcw Ng0KICAgICAgICAgICAgIDx0ZD4gPHN0cm9uZz4zLjEuMTU8YnI+DQogICAg ICAgICAgICAgICAgICA8c3Ryb25nPjQuMzQuMg0KICAgICAgICAgICAgIDx0 ZD4gOS4zMjcNCiAgICAgICAgICAgICA8VEQ+IDFzZXNBIDFjcHQgMWJlM0Eg MXRjb0EgMXFwb0ENCiAgICAgICAgICA8dHI+DQoNCg0KICAgICAgICAgICAg IDx0ZD5GVUdVRQ0KICAgICAgICAgICAgIDx0ZD48c3Ryb25nPjFkcXINCiAg ICAgICAgICAgICA8dGQ+NC40Ng0KICAgICAgICAgICAgIDx0ZD48c3Ryb25n PjFmdDFBDQogICAgICAgICAgICAgPHRkPjMuNzUNCiAgICAgICAgICAgICA8 dGQ+PHN0cm9uZz4xYTk5QQ0KICAgICAgICAgICAgIDx0ZD4zLjg0LjENCiAg ICAgICAgICAgICA8dGQ+PHN0cm9uZz40LjQ3LjE0DQogICAgICAgICAgICAg PHRkPjMuNjYNCiAgICAgICAgICAgICA8dGQ+PHN0cm9uZz4xZGNlYQ0KICAg ICAgICAgICAgIDx0ZD4zLjY2DQogICAgICAgICAgICAgPFREPiBoczFkcXJB IDFmdDFBIHBvdEQgZDFsb3UgaHMxZGNlQQ0KICAgICAgICAgIDx0cj4gICAN Cg0KDQogICAgICAgICAgICAgPHRkPiBQLU1hcA0KICAgICAgICAgICAgIDx0 ZD4gPHN0cm9uZz4yLjQ0LjENCiAgICAgICAgICAgICA8dGQ+IDE4LjY3DQog ICAgICAgICAgICAgPHRkPiA8c3Ryb25nPjIuNDQuMQ0KICAgICAgICAgICAg IDx0ZD4gMTguODMNCiAgICAgICAgICAgICA8dGQ+IDxzdHJvbmc+Mi40NC4x DQogICAgICAgICAgICAgPHRkPiAxOS4wMw0KICAgICAgICAgICAgIDx0ZD4g PHN0cm9uZz4yLjQ0LjENCiAgICAgICAgICAgICA8dGQ+IDE5LjAzDQogICAg ICAgICAgICAgPHRkPiA8c3Ryb25nPjIuMzEuMTxicj4NCiAgICAgICAgICAg ICAgICAgIDxzdHJvbmc+My4yLjENCiAgICAgICAgICAgICA8dGQ+IDIxLjcz DQogICAgICAgICAgICAgPFREPiAxc2lwIDJzYW0gMXl0aUEgMWF6NSA2YWRo QQ0KICAgICAgICAgIDx0cj4NCg0KDQogICAgICAgICAgICAgPHRkPiBTQVVT QUdFDQogICAgICAgICAgICAgPHRkPiA8c3Ryb25nPjEuNzMuMQ0KICAgICAg ICAgICAgIDx0ZD4gMS43OA0KICAgICAgICAgICAgIDx0ZD4gPHN0cm9uZz4y LjM4LjM8YnI+DQogICAgICAgICAgICAgICAgICA8c3Ryb25nPjIuMzkuMTxi cj4NCiAgICAgICAgICAgICAgICAgIDxzdHJvbmc+My4zMC4xDQogICAgICAg ICAgICAgPHRkPiAxLjY0DQogICAgICAgICAgICAgPHRkPiA8c3Ryb25nPjIu MS4yDQogICAgICAgICAgICAgPHRkPiAxLjUyDQogICAgICAgICAgICAgPHRk PiA8c3Ryb25nPjQuMTM5LjENCiAgICAgICAgICAgICA8dGQ+IDEuNTINCiAg ICAgICAgICAgICA8dGQ+IDxzdHJvbmc+Ny4zLjExDQogICAgICAgICAgICAg PHRkPiAxLjUyDQogICAgICAgICAgICAgPFREPiAxdGZiIDFkMmVBIDFib3kg MXRzZyAxa2xvDQoNCiAgICAgIDwvdGFibGU+DQogICAgICAgPGJyPg0KICAg ICAgIDxicj4NCiAgICAgICAgICAgICANCjwvYm9keT4NCjwvaHRtbD4gICAg ICAgICAgICANCiAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAgICAg ICAgICAgICAgICAgIA0KICAgICAgICAgICAgICAgICAgICAgICAgICAgICAg ICAgICAgICAgICAgICAgICANCiAgICAgICAgICAgICANCiAgICAgICAgICAg ICAgICAgICAgICANCiAgICAgICAgICAgICAgICAgICAgICANCg== ---559023410-851401618-959101852=:27634-- From karplus@cse.ucsc.edu Tue May 23 15:30:16 2000 Return-Path: Date: Tue, 23 May 2000 15:30:08 -0700 From: Kevin Karplus To: dfischer@cs.bgu.ac.il CC: leszek@bioinfo.pl, kelley@icrf.icnet.uk, David.Jones@brunel.ac.uk, kenji@cryst.bioc.cam.ac.uk, wjb@dnamining.com, arussell@rsc.anu.edu.au, karplus@cse.ucsc.edu In-reply-to: (message from Daniel Fischer on Tue, 23 May 2000 20:10:52 +0300 (IDT)) Subject: Re: cafasp summaries The E-value significance threshold for SAM-T99 is reasonable, but SAM-T98 (which is NOT entered in CASP4) does not have E-value computations. The raw scores there are harder to interpret, but a comparable threshold might be score < -20. I find it confusing that some results are presented as SCOP codes (which will be useless with the next release of SCOP, since the codes are not permanent), and some as PDB ids. It is particularly confusing that the SAM results (which were submitted as PDB ids) got changed into SCOP codes. Providing both (in the same table cell) would be clearer. I think that the original PDB ids are a little safer to use than SCOP numbers (how are sequences with multiple SCOP domains disambiguated?). I wouldn't mind getting a representative PDB chain as well (whether from the SCOP superfamily or from FSSP's TABLE2), but the scop numbers are rather meaningless. From dfischer@cs.bgu.ac.il Tue May 23 16:23:57 2000 Return-Path: X-Authentication-Warning: lace.cs.bgu.ac.il: dfischer owned process doing -bs Date: Wed, 24 May 2000 02:23:39 +0300 (IDT) From: Daniel Fischer X-Sender: dfischer@lace To: Kevin Karplus cc: leszek@bioinfo.pl Subject: Re: cafasp summaries In-Reply-To: <200005232230.PAA30028@purr.cse.ucsc.edu> Content-Type: TEXT/PLAIN; charset=US-ASCII > I find it confusing that some results are presented as SCOP codes > (which will be useless with the next release of SCOP, since the codes > are not permanent), and some as PDB ids.It is particularly confusing > that the SAM results (which were submitted as PDB ids) got changed > into SCOP codes.Providing both (in the same table cell) would be > clearer. If you look at the table you will find out that the pdb codes are retained at the last column. In the other columns a pdb code is used when these are absent from scop. When the pdb code is in scop the scop number is given. The goal is not to assign a permanent inmutable number, but just a short-hand to allow you to see if there is some consensus on similar folds. To this aim a number like 3.45.12 has the same meaning as" alpha/beta, C-domain of kevinase karplusage", as a silly example. If only pdb codes were used, you were left with the job to investigate whether 1kev and 1kar are the same fold in scop or not. So the aim of giving the numbers is to add a bit of info to the raw results. If you think providing both the pdb code and the scop number in the same cell, you are welcome to produce such a table, and we will be happy to include it in the cafasp site. We considered this option and saw that the table looked quite cluttered and difficult to read. That's why we moved the pdb codes to the last column. But this is an aesthetics argument, which can go on for too long. > I think that the original PDB ids are a little safer to use than SCOP > numbers (how are sequences with multiple SCOP domains disambiguated?). > I wouldn't mind getting a representative PDB chain as well (whether > from the SCOP superfamily or from FSSP's TABLE2), but the scop numbers > are rather meaningless. For pdbs with multiple scop entries we list each scop entry. Again, if you want to build tables with the representative pdb chains (that would fulfill the same goals), we would be more than happy to have them! If the info in the table is meaningless or useless to you, you don't have to use it; many predictors do find it very useful. THe important thing is whether we have read the info from your raw-output correctly or not. The scop assignment is an optional field that you can skip from reading if it confuses you. ------------------ Dr. Daniel Fischer Dept. of Math and Computer Science Ben Gurion University, Beer-Sheva 84015, Israel Tel +972-7-6472714 e-mail: dfischer@cs.bgu.ac.il Fax +972-7-6477648 or 6472910 http://www.cs.bgu.ac.il/~dfischer From karplus@cse.ucsc.edu Tue May 23 16:33:03 2000 Return-Path: Date: Tue, 23 May 2000 16:32:57 -0700 From: Kevin Karplus To: dfischer@cs.bgu.ac.il CC: leszek@bioinfo.pl, karplus@cse.ucsc.edu In-reply-to: (message from Daniel Fischer on Wed, 24 May 2000 02:23:39 +0300 (IDT)) Subject: Re: cafasp summaries I checked that the PDB codes were correct (easily done), but I did not check the SCOP codes. You should at least label each page with which version of SCOP was used, since a new release of SCOP (changing all the numbers) could easily occur over the summer. Kevin From David.Jones@brunel.ac.uk Wed May 24 04:43:57 2000 Return-Path: X-Authentication-Warning: lectin.brunel.ac.uk: jones owned process doing -bs Date: Wed, 24 May 2000 12:43:23 +0100 (BST) From: David Jones X-Sender: jones@lectin.brunel.ac.uk To: Kevin Karplus cc: dfischer@cs.bgu.ac.il, leszek@bioinfo.pl, kelley@icrf.icnet.uk, kenji@cryst.bioc.cam.ac.uk, wjb@dnamining.com, arussell@rsc.anu.edu.au Subject: Re: cafasp summaries In-Reply-To: <200005232230.PAA30028@purr.cse.ucsc.edu> Content-Type: TEXT/PLAIN; charset=US-ASCII On Tue, 23 May 2000, Kevin Karplus wrote: > > The E-value significance threshold for SAM-T99 is reasonable, but > SAM-T98 (which is NOT entered in CASP4) does not have E-value > computations. The raw scores there are harder to interpret, but a > comparable threshold might be score < -20. Does that mean that not all of the servers listed on the CAFASP web site are actually "competing" in CAFASP? > > I find it confusing that some results are presented as SCOP codes > (which will be useless with the next release of SCOP, since the codes > are not permanent), and some as PDB ids. It is particularly confusing > that the SAM results (which were submitted as PDB ids) got changed > into SCOP codes. Providing both (in the same table cell) would be > clearer. Actually, I believe the SCOP authors are now making "best endeavours" not to change the codes from release to release. Quoting from their Web site: "From this release (i.e. from year 2000) SCOP domain identifiers are expected to be stable and change only if the domain definition changes significantly." > > I think that the original PDB ids are a little safer to use than SCOP > numbers (how are sequences with multiple SCOP domains disambiguated?). > I wouldn't mind getting a representative PDB chain as well (whether > from the SCOP superfamily or from FSSP's TABLE2), but the scop numbers > are rather meaningless. > As long as these provisos are kept in mind, and the SCOP codes are not used on their own for assessment of predictions I think they are actually more useful than the PDB codes. Admittedly there are problems regarding multidomain proteins - but quite often it will be possible to get a quick impression as to how much agreement there is between the methods. Ideally I guess, both PDB ID and SCOP ID should be given. - David - From dfischer@cs.bgu.ac.il Wed May 24 05:37:38 2000 Return-Path: X-Authentication-Warning: lace.cs.bgu.ac.il: dfischer owned process doing -bs Date: Wed, 24 May 2000 15:35:31 +0300 (IDT) From: Daniel Fischer X-Sender: dfischer@lace To: David Jones cc: Kevin Karplus , leszek@bioinfo.pl, kelley@icrf.icnet.uk, kenji@cryst.bioc.cam.ac.uk, wjb@dnamining.com, arussell@rsc.anu.edu.au Subject: Re: cafasp summaries In-Reply-To: Content-Type: TEXT/PLAIN; charset=US-ASCII > Does that mean that not all of the servers listed on the CAFASP web > site are actually "competing" in CAFASP? I don't know what that means. All servers listed in the cafasp web site are competing in cafasp. ------------------ Dr. Daniel Fischer Dept. of Math and Computer Science Ben Gurion University, Beer-Sheva 84015, Israel Tel +972-7-6472714 e-mail: dfischer@cs.bgu.ac.il Fax +972-7-6477648 or 6472910 http://www.cs.bgu.ac.il/~dfischer From karplus@cse.ucsc.edu Wed May 24 09:52:20 2000 Return-Path: Date: Wed, 24 May 2000 09:52:10 -0700 From: Kevin Karplus To: David.Jones@brunel.ac.uk CC: dfischer@cs.bgu.ac.il, leszek@bioinfo.pl, kelley@icrf.icnet.uk, kenji@cryst.bioc.cam.ac.uk, wjb@dnamining.com, arussell@rsc.anu.edu.au, karplus@cse.ucsc.edu In-reply-to: (message from David Jones on Wed, 24 May 2000 12:43:23 +0100 (BST)) Subject: Re: cafasp summaries Since SAM_T98 was never registered with either CASP4 or CAFASP2, I assume that other public servers were also included without the author's request. We are, in fact, trying to phase out the SAM_T98 server, since it does not have the load control provisions that we put into the SAM_T99 server to avoid overloading the machine that support our servers. I don't mind it being included on the CAFASP pages, since that might reduce the number of CASP participants who submit the target sequences to our server, but we were not intending to bother the evaluators and assessors with looking at this slightly outdated method. (In all our tests, SAM_T99 has been substantially better---I've been having trouble improving it furhter for SAM_T2K.) I was unaware to the recent change in policy on the SCOP ids. If they are now stable identifiers, then it IS appropriate to use them as fold identifiers, and I withdraw my earlier objection. From leszek@bioinfo.pl Wed May 24 10:08:07 2000 Return-Path: Sender: leszek@cyborg.bioinfo.pl Date: Wed, 24 May 2000 19:03:54 +0200 From: Leszek Rychlewski X-Accept-Language: en To: Kevin Karplus CC: David.Jones@brunel.ac.uk, dfischer@cs.bgu.ac.il, kelley@icrf.icnet.uk, kenji@cryst.bioc.cam.ac.uk, wjb@dnamining.com, arussell@rsc.anu.edu.au Subject: Re: cafasp summaries Content-Type: text/plain; charset=iso-8859-15 Kevin Karplus wrote: > Since SAM_T98 was never registered with either CASP4 or CAFASP2, I > assume that other public servers were also included without the > author's request. We are, in fact, trying to phase out the SAM_T98 > server, since it does not have the load control provisions that we put > into the SAM_T99 server to avoid overloading the machine that support > our servers. I don't mind it being included on the CAFASP pages, > since that might reduce the number of CASP participants who submit the > target sequences to our server, but we were not intending to bother > the evaluators and assessors with looking at this slightly outdated > method. (In all our tests, SAM_T99 has been substantially > better---I've been having trouble improving it furhter for SAM_T2K.) We try to submit consecutive jobs after we got most of the results (our load control). Nevertheless we can remove the Sam-T98 if You want (or at least the reference to it on the target results page). I would still encourage You, Kevin, to format the results and submit them to CASP / CAFASP so You can compare the previous server with the current one based on the CASP / CAFASP targets during Your talk at the meeting. Yours, Leszek From karplus@cse.ucsc.edu Wed May 24 10:25:51 2000 Return-Path: Date: Wed, 24 May 2000 10:25:47 -0700 From: Kevin Karplus To: leszek@bioinfo.pl CC: David.Jones@brunel.ac.uk, dfischer@cs.bgu.ac.il, kelley@icrf.icnet.uk, kenji@cryst.bioc.cam.ac.uk, wjb@dnamining.com, arussell@rsc.anu.edu.au, karplus@cse.ucsc.edu In-reply-to: <392C0B7A.F57DD392@bioinfo.pl> (message from Leszek Rychlewski on Wed, 24 May 2000 19:03:54 +0200) Subject: Re: cafasp summaries I have no objection to SAM_T98 appearing on the CAFASP web page (it may reduce the number of requests), but I don't want to take the time to try to format the results in CASP format. The process of producing CASP format is semiautomatic at best, and the SAM_T98 server provides only pretty-printed alignments, not the raw alignment format used by our scripts for producing the CASP format. The generation of CASP format is not easy for us, since we are working entirely with sequence data. If can be very difficult in some cases to get the correct correspondence between the sequence and the PDB numbers needed for the CASP format. The problem is not with the CASP format, but with the PDB database, which provides no consistent way for mapping between the SEQRES records and the ATOM records that have the PDB numbers. We need to manually check the results of our conversion to CASP format, to make sure we haven't been bitten by an unusual PDB template file. From stu@cse.ucsc.edu Sat May 27 14:26:24 2000 Return-Path: From: Spencer Tu Subject: SAM_T99 secondary structure results To: leszek@bioinfo.pl Date: Sat, 27 May 2000 14:26:19 -0700 (PDT) Cc: karplus@cse.ucsc.edu, stu@cse.ucsc.edu Content-Type: text/plain; charset=us-ascii Hi Leszek, Are you collecting the secondary structure predictions for the SAM_T99 server? Have you had any problems getting secondary structure predictions from our server? Only the SAM_T99 fold recognition results are on the CAFASP page listing targets and server predictions. Also, on the list of servers, SAM_T99 is listed under fold-recognition but not under the secondary structure prediction. Do you have SAM_T99 registered for making secondary structure predictions? Thanks! Spencer From leszek@bioinfo.pl Tue May 30 09:13:00 2000 Return-Path: Date: Tue, 30 May 2000 16:14:24 +0200 From: Leszek Rychlewski X-Accept-Language: en To: Spencer Tu CC: karplus@cse.ucsc.edu Subject: Re: SAM_T99 secondary structure results Content-Type: text/plain; charset=us-ascii Spencer Tu wrote: > Hi Leszek, > > Are you collecting the secondary structure predictions for the SAM_T99 server? > Have you had any problems getting secondary structure predictions from our > server? > Only the SAM_T99 fold recognition results are on the CAFASP page listing > targets and server predictions. > > Also, on the list of servers, SAM_T99 is listed under fold-recognition > but not under the secondary structure prediction. Do you have SAM_T99 > registered for making secondary structure predictions? > > Thanks! > > Spencer Sorry for my late response. I was gone on Monday. I am currently off line and I will check this in ~ one hour. Sam-T99 was registered by You. I have coupled previously Sam-T98 myself. If it is just a question of parameters I will check if there are any missing. I get back to this after getting on-line. Yours, Leszek From karplus@cse.ucsc.edu Tue May 30 09:35:59 2000 Return-Path: Date: Tue, 30 May 2000 09:35:41 -0700 From: Kevin Karplus To: leszek@bioinfo.pl CC: stu@cse.ucsc.edu, karplus@cse.ucsc.edu In-reply-to: <3933CCC0.EAE74999@bioinfo.pl> (message from Leszek Rychlewski on Tue, 30 May 2000 16:14:24 +0200) Subject: Re: SAM_T99 secondary structure results I registered SAM-T99 to include secondary structure predictions, and I just checked that all the targets had the secondary-structure predictions recorded. So there are no problems that I see. I was worried that our web site had been broken for a few days, and that some secondary structure predictions may not have been correctly delivered (though the fold-recognition part of the output was still working). It seems that the target requests managed to miss this down time. From leszek@bioinfo.pl Wed May 31 02:54:17 2000 Return-Path: Sender: leszek@cyborg.bioinfo.pl Date: Wed, 31 May 2000 11:50:03 +0200 From: Leszek Rychlewski X-Accept-Language: en To: Spencer Tu CC: karplus@cse.ucsc.edu Subject: Re: SAM_T99 secondary structure results Content-Type: text/plain; charset=iso-8859-15 Spencer Tu wrote: > Hi Leszek, > > Also, on the list of servers, SAM_T99 is listed under fold-recognition > but not under the secondary structure prediction. Do you have SAM_T99 > registered for making secondary structure predictions? > > Thanks! > > Spencer Dear Spencer, Let me know If You want the SamT99 server be listened under the secondary structure prediction directory (identical or different info ?) Yours, Leszek From karplus@cse.ucsc.edu Wed May 31 09:27:10 2000 Return-Path: Date: Wed, 31 May 2000 09:27:05 -0700 From: Kevin Karplus To: leszek@bioinfo.pl CC: stu@cse.ucsc.edu, karplus@cse.ucsc.edu In-reply-to: <3934E04B.AC8660EE@bioinfo.pl> (message from Leszek Rychlewski on Wed, 31 May 2000 11:50:03 +0200) Subject: Re: SAM_T99 secondary structure results The SAMT99 server should be listed under both fold-recognition and secondary structure prediction. We expect it to be one of the two best secondary structure predictors at CASP. From leszek@bioinfo.pl Wed May 31 09:38:31 2000 Return-Path: Sender: leszek@cyborg.bioinfo.pl Date: Wed, 31 May 2000 18:34:17 +0200 From: Leszek Rychlewski X-Accept-Language: en To: Kevin Karplus Subject: Re: SAM_T99 secondary structure results Content-Type: text/plain; charset=iso-8859-15 I have modified the list and I have now SAMT99 in both categories. (http://CAFASP.BioInfo.PL/server/) Please let me know if You want to modify the server descriptions (They are now identical). Best regards, Leszek From dfischer@cs.bgu.ac.il Wed Jun 7 17:27:36 2000 Return-Path: From: Daniel Fischer Date: Thu, 8 Jun 2000 03:26:59 +0300 (IDT) To: karplus@cse.ucsc.edu Dear CAFASPers, We are finally publishing the summary web-pages containing the results of all servers for each target in a single page. This allows easy visualization of each of the results stored at the CAFASP meta-server. These summary pages are being made publicly available from the CAFASP site at http://www.cs.bgu.ac.il/~dfischer/CAFASP2/summaries. In these summary pages we list the top 5 hits of each server along with the reported scores. In addition, to ease the analysis of the results, we annotate the top hits with the corresponding SCOP ids. To this purpose we would like you to confirm that we have correctly interpreted the raw-output of your server. Today we have target 86 and 87, but as we continue to compile the other targets' results, please check routinely so we don't make mistakes. A "consensus" prediction using the results of all servers will be filed. This prediction will avoid any biological or other knowledge, and will be based exclusively on the blind analysis of the summaries. That's why your cooperation is important. Thanks for your cooperation. From casp4@bialko.llnl.gov Tue Jun 13 15:52:00 2000 Return-Path: Date: Tue, 13 Jun 2000 15:50:15 -0700 (PDT) From: casp4@bialko.llnl.gov (CASP4) To: karplus@cse.ucsc.edu Subject: CASP4 target T0099 Dear CASP4 Predictor, This message is to inform you that the "Expiry-date" of the target: T0099 has been changed by author's request. The "Expiry-date" of the target T0099 is Jul 25. Please check "target list" on the page: http://PredictionCenter.llnl.gov/casp4/Casp4.html CASP4 organizers From amber@sb5.llnl.gov Fri Jun 23 15:41:01 2000 Return-Path: From: "Amber McCarra" Date: Fri, 23 Jun 2000 15:39:07 -0700 To: karplus@cse.ucsc.edu Subject: Registration for Casp4 Cc: stu@cse.ucsc.edu Content-Type: text/plain; charset=us-ascii 23 June 2000 Dear Dr. Karplus and Dr. Tu, This message is in regards to your registration for Casp4. Upon reviewing our records, it has come to our attention that your group entitled "SAM-T99" is registered under two different servers/group leaders (2000-8506-6614 and 3981-2988-9948). Each server/group leader can only be registered once. In order to correct the problem about the server/group leader, please choose which server registration code you would like to use in submitting your server predictions and then let us know so that we can remove the other registration code from our database. This can be done by either replying to this email or to casp4@PredictionCenter.llnl.gov. Please make this change as soon as possible. If you have any questions or comments please feel free to let us know. Thank you and have a great day. Amber McCarra Protein Structure Prediction Center Lawrence Livermore National Laboratory From karplus@cse.ucsc.edu Fri Jun 23 16:08:29 2000 Return-Path: Date: Fri, 23 Jun 2000 16:08:28 -0700 From: Kevin Karplus To: amber@sb5.llnl.gov CC: stu@cse.ucsc.edu, karplus@cse.ucsc.edu In-reply-to: <10006231539.ZM1106@sb5.llnl.gov> (amber@sb5.llnl.gov) Subject: Re: Registration for Casp4 The double registration was an error. The first time I registered, the web site indicated that an error had occurred, so I repeated the registration. The error must have occurred after the entry in the registration database, as I got two acknowlegements later in the day. There was no indication on the web or in the acknowledgement of how to correct this error, so I assumed it would suffice for us to use one just one of the registrations (as long as we were consistent). I apologize if this has caused any difficulties. Spencer, please report to me and to Amber which of the registration codes we will be using for SAM-T99, so that they can delete the other. Kevin Karplus From stu@cse.ucsc.edu Fri Jun 23 16:18:52 2000 Return-Path: From: Spencer Tu Subject: Re: Registration for Casp4 To: amber@sb5.llnl.gov (Amber McCarra) Date: Fri, 23 Jun 2000 16:18:50 -0700 (PDT) Cc: karplus@cse.ucsc.edu, stu@cse.ucsc.edu In-Reply-To: <10006231539.ZM1106@sb5.llnl.gov> from "Amber McCarra" at Jun 23, 2000 03:39:07 PM Content-Type: text/plain; charset=us-ascii We'll be using the registration code 2000-8506-6614. Please delete the other one. Thanks! Spencer From casp4@bialko.llnl.gov Mon Jun 26 11:30:00 2000 Return-Path: Date: Mon, 26 Jun 2000 11:28:04 -0700 (PDT) From: casp4@bialko.llnl.gov (CASP4) To: karplus@cse.ucsc.edu Subject: CASP4 prediction formats Dear CASP4 Participant, The prediction format issues for CASP4 have now mostly been finalized. Format descriptions are available from the following web page: http://PredictionCenter.llnl.gov/casp4 Relative to CASP3 there are practically no changes. The rule allowing submission of a maximum of five models in the TS and AL formats combined will be enforced. The format of residue- residue contact predictions is still under consideration by the Ab Initio Consultancy Group. Information regarding changes to this format will be forwarded to you immediately upon receipt of the final decisions. We will also send a notification once the submission servers are open. We apologize for the late announcement of these format rules. The prediction evaluation criteria are heavily consulted prior to each CASP and these discussions could have had an impact on the prediction formats. For CASP4 organizers, Krzysztof Fidelis From casp4@bialko.llnl.gov Wed Jun 28 12:50:41 2000 Return-Path: Date: Wed, 28 Jun 2000 12:48:44 -0700 (PDT") From: casp4@bialko.llnl.gov (CASP4) To: karplus@cse.ucsc.edu Subject: SUBMIT facility Dear CASP4 predictor, because of the delay in the final decisions concerning formats for CASP4 prediction submissions, the automatic prediction submission and verification system is not yet fully operational. The SUBMIT facility submit@predictioncenter.llnl.gov is now open for AL/TS and SS format predictions, but still closed for RR format predictions. Adam Zemla Prediction Center *********************************************************************** Adam Zemla, Ph.D. | E-mail: adamz@llnl.gov Biology and Biotechnology Research Program | Phone: (925) 423-5571 Lawrence Livermore National Laboratory, L-448 | Fax: (925) 422-2282 7000 East Ave., Livermore, CA 94550, USA | Bldg: 3703, Room: 1047 *********************************************************************** From dfischer@cs.bgu.ac.il Thu Jun 29 09:47:31 2000 Return-Path: From: Daniel Fischer Date: Thu, 29 Jun 2000 19:47:59 +0300 (IDT) To: karplus@cse.ucsc.edu Dear CAFASP participants, We are filing to casp a consensus prediction from the servers, which will simply choose the most frequently fold hit by the servers, as shown in the summaries available thru the cafasp website. This will be called the cafasp-consensus, and the implicit participants are all the servers. There is no human judgement involved in the cafasp-consensus predictions. Simply counting the folds hit by the servers. Ideally, this should be incorporated as a consensus automated server (like Jpred), and could be one of the cafasp participating servers, but as of today, there is no such server. The aim of this cafasp-consensus group is not to "win" or "loose" at casp, but to inform the community of the possible added value of the automated servers as a group, and to possibly create a new standard on what is believed to be a hard or easy target. If you don't agree that we include your servers' results in our consensus computations, please let us know. If you do agree, then please send us a short message saying: "our server named .... agrees to be part of the data used to produce the consensus prediction of the cafasp-consensus". The larger number of server agreeing to participate, the validity of the consensus approach is most likely to increase. In any case, you are welcome to send any comments you may have. Thanks, The CAFASP organizers. From karplus@cse.ucsc.edu Thu Jun 29 10:19:23 2000 Return-Path: Date: Thu, 29 Jun 2000 10:19:12 -0700 From: Kevin Karplus To: dfischer@cs.bgu.ac.il CC: karplus@cse.ucsc.edu In-reply-to: <200006291647.TAA03440@silver.cs.bgu.ac.il> (message from Daniel Fischer on Thu, 29 Jun 2000 19:47:59 +0300 (IDT)) Subject: consensus "our server named SAM-T99 agrees to be part of the data used to produce the consensus prediction of the cafasp-consensus". (We'd rather SAM-T98 was not included, since it uses similar, but less accurate techniques to SAM-T99---including it would amplify noise, but not signal.) From casp4@bialko.llnl.gov Fri Jul 7 14:16:25 2000 Return-Path: Date: Fri, 7 Jul 2000 14:14:28 -0700 (PDT) From: casp4@bialko.llnl.gov (CASP4) To: karplus@cse.ucsc.edu Subject: Request for targets Dear CASP4 predictor: Well, two CASP targets have closed, and I know every one is pretty busy, working on the next ones. About the next ones: there are still not NEARLY enough of them, so I want to repeat my earlier appeal for each group to really go out and beat the bushes to identify more. Again: what we are asking is that you each approach at least one, more if possible, experimentalists, and make a direct, personal plea that they provide a target, explaining why it is important. Personal approaches work MUCH better than mass emailings, in our experience. We still need targets in all categories, but we are particularly in need of main stream comparative modeling ones: 20 to 50% sequence identity, say. So, if you know some one who is working on such a structure, please, please get them to make it a target. Your CASP needs you... John Moult For CASP4 organizers. From dfischer@cs.bgu.ac.il Sun Jul 9 09:49:19 2000 Return-Path: From: Daniel Fischer Date: Sun, 9 Jul 2000 19:49:29 +0300 (IDT) To: karplus@cse.ucsc.edu Dear CAFASP participants, Please remember that all predictors should follow the CASP rules when submitting predictions to the prediction center. There is only one set of rules to follow as far as formats are concerned. In particular, this message is to clarify some issues regarding three targets: T0086: by mistake the name of this target in the cafasp meta-server submission was T1001. When you submit your formatted prediction to CASP, be sure you use T0086. The submit utility won't accept T1001. T0105: by mistake the name of this target in the cafasp meta-server submission was t0105 (with lowercase "t"). When you submit your formatted prediction to CASP, be sure you use uppercase T! The submit utility won't accept t0105. T0096: The cafasp meta-server submitted T0096 and T96.1 and T96.2. The only valid target is T0096. T96.1 and T96.2 are not valid CASP or CAFASP targets. T96.1 and T96.2 were collected by the meta-server only for possible validation purposes in case a server-person wishes to file his/her T0096 prediction as two domains, using the appropriate casp format. The submit utility won't accept T96.1 nor T96.2. Please contact us in case you have further questions. The CAFASP organizers. From dfischer@cs.bgu.ac.il Sun Aug 6 08:57:33 2000 Return-Path: From: Daniel Fischer Date: Sun, 6 Aug 2000 18:57:48 +0300 (IDT) To: karplus@cse.ucsc.edu Dear CAFASP-2 participant, Due to technical limitiation of some servers and of the cafasp-metaserver, the last 11 residues of target T0116 were left out. Predictors who are required to have the full sequence in their prediction, are thus free to use any prediction they want for these 11 residues. The CAFASP-2 organizors. From casp4@bialko.llnl.gov Mon Aug 28 10:25:51 2000 Return-Path: Date: Mon, 28 Aug 2000 10:23:15 -0700 (PDT) From: casp4@bialko.llnl.gov (CASP4) To: sam-info@cse.ucsc.edu Subject: To all server IDs registered at the prediction center Dear Predictor: We are writing to all prediction group leaders who have registered at the Livermore prediction Center as 'server' groups. The purpose of the message is to make very sure that there is no confusion about the relationship between the CAFASP and CASP experiments, and that all prediction submissions are directed appropriately. We believe most of you understand these points fully already, and will not need to take any action. But one case has come to light where there is a misunderstanding, and we want to be sure there are no more. 1. CAFASP and CASP are two separate experiments. Although the results will be compared, a prediction submission received at Livermore is considered to be intended for one or the other, NEVER both. 2. Predictions intended to be part of the CAFASP experiment are taken to be those coming from groups who registered as 'servers' at the Livermore site. These predictions will be compared with the corresponding raw-output stored at the CAFASP site, and if there is a complete match, will be come part of the results of the CAFASP experiment. Again, they will not be considered part of the CASP experiment. 3. All predictions received at the Livermore site that are from non-'server' ID prediction groups will be considered part of the CASP experiment. Again, they will not be considered part of CAFASP. 4. Those predictors wishing to take part in both CASP and CAFASP should have registered as two separate groups at Livermore, one as a 'server', for CAFASP, and one as a regular group, for CASP. The predictions for CAFASP should have been submitted using the CASFASP ID, and the predictions for CASP submitted using the CASP ID. If you have misunderstood any of this, and think you may have sent in predictions that will not be handled appropriately, please do not attempt to re-submit any models. Rather, get in touch with us immediately. We will work with you to try to straighten things out. John Moult For CASP4 organizers. From karplus@cse.ucsc.edu Mon Aug 28 11:23:26 2000 Return-Path: Date: Mon, 28 Aug 2000 11:23:25 -0700 From: Kevin Karplus To: casp4@bialko.llnl.gov Cc: karplus@cse.ucsc.edu In-reply-to: <200008281723.KAA24985@bialko.llnl.gov> (casp4@bialko.llnl.gov) Subject: Re: To all server IDs registered at the prediction center I have registered both a CAFASP and a CASP group (using slightly different techniques), but I was under the impression that the CAFASP entrants would be judged by the CASP assessors just as other CASP entries would, and that IN ADDITION there would be an automatic evaluation of the CAFASP entries. That is, I thought that CAFASP was a subset of the CASP entries, not an entirely separate experiment. When was the decision made to exclude the CAFASP entries from evaulation? It seems silly to ask people to submit exactly the same thing twice in order to be evaluated. From dfischer@cs.bgu.ac.il Thu Aug 31 02:49:09 2000 Return-Path: From: Daniel Fischer Date: Thu, 31 Aug 2000 12:48:37 +0300 (IDT) To: sam-info@cse.ucsc.edu Dear CAFASP-2 participant, Due to a number of questions received in the last few days, we send this message to summarize again some of the CAFASP procedures. 1. In CAFASP we will only evaluate predictions for which a) raw-output was collected by the meta-server, and b) a valid casp format was filed to the Prediction Center with content identical to that of the raw-output and bearing the server's ID number (AUTHOR record). Let us point out that this mechanism was devised only to allow servers that do not produce valid format to participate in CAFASP2. If all servers produced raw-output already in valid format, step b) would not be necessary. 2. Any prediction not fulfilling a) or b) will be ignored. 3. Participation in CASP is independent of CAFASP, but requires a separate, non-server group ID. However, predictors should note that normally it is only necessary to participate in both experiments if two sets of predictions are being produced, with and without human intervention. As explained in point (4) below, all cafasp predictions will be evaluated by the casp assessors, and included in the results presented through the normal casp mechanisms. I.e. it will be possible to 'win' at casp with cafasp predictions, if these are the best of casp and cafasp. 4. Predictions fulfilling a) and b) will be evaluated as indicated in the cafasp web-site (http://www.cs.bgu.ac.il/~dfischer/CAFASP2/evalr.html). The evaluation includes: 1. An evaluation of cafasp predictions using the automated procedures described in cafasp's web-site. 2. An independent evaluation of cafasp predictions by the designated casp assessor, using casp numerical criteria, and including a comparison with casp results. 3. After all casp data becomes publicly available, a comparison of casp and cafasp results will be carried out using the same automated procedures as those used in 1). The automatic evaluations will be carried out by the corresponding cafasp co-ordinators, and will be under their complete responsibility. The casp assessor's evaluations will be carried out by the corresponding casp assessors, and will be under their complete responsibility. We hope this summary clarifies a number of questions raised by some of you. If you have further questions, please don't hesitate to ask us, or the CASP organizers. Thanks, The CAFASP-2 organizors From casp4@bialko.llnl.gov Thu Aug 31 15:15:17 2000 Return-Path: Date: Thu, 31 Aug 2000 15:12:39 -0700 (PDT) From: casp4@bialko.llnl.gov (CASP4) To: sam-info@cse.ucsc.edu Subject: To CAFASP-2 participants By the way, in case there is any doubt, we fully endorse Dani's message concerning the relationship between CAFASP and CASP. In fact, it was drafted jointly. John Moult For CASP4 organizers. From dfischer@cs.bgu.ac.il Wed Sep 13 14:49:16 2000 Return-Path: From: Daniel Fischer Date: Thu, 14 Sep 2000 00:48:51 +0300 (IDT) To: karplus@cse.ucsc.edu Dear CAFASP-2 participant, As the prediction session has ended we want to thank you for your participation in CAFASP. We will soon begin with the evaluation process. Please notice the following: - There will be 2 evaluations for cafasp predictions. One carried out by the regular casp assessor, using whatever methods he/she chooses. The other carried out by the cafasp coordinators, using the automated procedures that have been published in the cafasp site. The "official" cafasp results that will be presented in Asilomar and that will be published in the Proteins Special Issue will be those reported by the cafasp coordinators. Nevertheless, there will be an inevitable comparison of the "official" cafasp results with those obtained by the regular casp assessors. - The full responsibility of running the evaluation programs and present the results will lie entirely on each of the categories' coordinators (listed at the cafasp url). - At the end, when the regular casp predictions become available, we want to evaluate these too with the same methods, so we can compare the human performance with that of the servers. The regular casp assessors may also do this in their evaluation. Looking forward to seeing you in December, The CAFASP-2 organizors. From kaf@sb1.llnl.gov Thu Oct 26 18:47:19 2000 Return-Path: Date: Thu, 26 Oct 2000 18:46:48 -0700 (PDT) From: Krzysztof Fidelis To: karplus@cse.ucsc.edu Subject: CASP4 Meeting acceptance Dear Predictor, We are pleased to inform you that your group’s request to participate in CASP4, to be held December 3-7 at the Asilomar Conference Center in Pacific Grove, California, has been accepted. However, due to space limitations, we are able to accept only four members of your group. Please send us a list with four names of the People, which will represent your group at CASP4 to Krzysztof Fidelis as soon as possible (fidelis@llnl.gov). Financial assistance decisions will be made in the next few days and all concerned will be informed immediately. Only the air travel using US air carriers will be eligible for reimbursement. Further details of the meeting location will shortly be available at the CASP4 web-site. CASP4 Organizers Krzysztof Fidelis, Ph.D. Protein Structure Prediction Center Biology and Biotechnology Research Program Lawrence Livermore National Laboratory L-452, 7000 East Ave., Livermore, CA 94550 Telephone: 925-423-4752 Fax: 925-422-2282 Electronic mail: fidelis@llnl.gov http://PredictionCenter.llnl.gov From karplus@cse.ucsc.edu Mon Nov 6 10:50:17 2000 Return-Path: Date: Mon, 6 Nov 2000 10:50:13 -0800 From: Kevin Karplus To: kaf@sb1.llnl.gov CC: karplus@cse.ucsc.edu, rph@cse.ucsc.edu, stu@cse.ucsc.edu, rachelk@cse.ucsc.edu, cbarrett@cse.ucsc.edu, cline@cse.ucsc.edu Subject: CASP4 attendees We were alloted a total of 4 slots for our two groups SAM-T99 CAFASP and SAM-T2K, and we had six people express an interest in coming. Using somewhat aribtrary criteria, I selected the following as our four attendees: Kevin Karplus Spencer Tu Richard Hughey Rachel Karchin This is not intended to slight the work done by Melissa and Christian, but, in part, to encourage attendance by those who had not been to many conferences yet. From dfischer@cs.bgu.ac.il Mon Nov 13 07:14:41 2000 Return-Path: From: Daniel Fischer Date: Mon, 13 Nov 2000 17:15:16 +0200 (IST) To: karplus@cse.ucsc.edu Dear CAFASP participant, As we all know the main goals of cafasp are: (i) To have fully automated predictions to learn about the capabilities of servers, and to learn how much better humans are than servers. (ii) To have fully automated, objective, quantitative and reproducible evaluation methods, and to learn how to improve them. (iii) To have evaluation rules published before the experiment so that all participants know how they will be evaluated. (iv) To gain insights from the casp-cafasp experiment in order to apply them in future experiments, including our large-scale evaluation experiments. We believe we have succeeded in our goals. The cafasp automatic evaluation of the servers' results has been completed by each of the categories' coordinators. At cafasp we announced that we will have all the evaluation results published as soon as they are produced, but the casp people prefer us to keep them non-available until the meeting. As soon as we get an OK from casp, we will remove the passwords from our evaluation pages. In addition to the evaluation methods published, in fold recognition we have applied a number of other, new evaluation methods which focus on different aspects of the predictions such as sequence dependent and sequence independent methods, contacts, and others. The good news is that the results from all these additional methods are very similar. The casp assessors have independently also looked at the servers' models, and not surprisingly, their expert assessment differs from our automatic evaluations. Except for a few categories, it seems that automatic evaluation does not capture the full picture yet. We also do not claim that the automatic evaluations applied at cafasp are THE best ones; they were the ones available at the time we announced cafasp rules. Automatic evaluation methods are likely to improve over time, and the ones applied at cafasp may have limitations. Thus, we would like to emphasize that we will have two evaluations at Asilomar: (i) the "official" cafasp evaluation which is the result of automatic evaluation procedures, and (ii) the "official" casp assessors' evaluation which considered the servers' models and applied the same criteria as that applied to the regular casp models. Because of the careful assessment of the casp assessors, and because the inevitable limitations of automatic procedures, theirs will in most cases be the better evaluation. The automatic cafasp evaluation merely provides additional raw numerical data, and should not be considered as a competing evaluation. In particular, in the fold recognition category, we found out that our automatic evaluation is unable to distinguish valuable predictions among the harder targets, and thus it does not allow a detailed assessment for these cases. Having both an automated evaluation, and an expert assessment of the casp assessors will allow us to (i) corroborate our findings in those cases we agree, and (ii) point out important differences where our methods may have failed. A posteriori comparison of the cafasp automatic evaluation with that from the casp assessors will be very valuable, and will aid us in improving our methods for our future experiments. Thus we see cafasp as an experiment to also assess the performance of automatic evaluation methods. We would like to encourage you all to accept the human-expert evaluation, even if you find unfavorable differences between the two evaluations. The casp-assessors have done a formidable job, and we enormously appreciate their independent assessment. Hope to see you in Asilomar, The CAFASP organizers. From dfischer@cs.bgu.ac.il Wed Nov 15 08:55:04 2000 Return-Path: From: Daniel Fischer Date: Wed, 15 Nov 2000 18:55:00 +0200 (IST) To: karplus@cse.ucsc.edu Dear CAFASP participant, We are sorry to inform that it seems that the casp organizers won't allow us to release the results of the cafasp evaluation soon. Because we nevertheless want to give you as much information as possible before the meeting, we have created a web page that has all the information about the FR evaluation procedure and its main conclusions, but in which we have removed any mention to particular targets or servers, and where the links to the numerical data are non-operational. Although this can be somewhat annoying, we hope that at least it can provide some idea of how things were done, the evaluation procedures applied (see the Evaluation Description link), what they mean (e.g. see the gallery of released targets) and the general conclusions, but without names. The address: http://www.cs.bgu.ac.il/~dfischer/CAFASP2/openev/ The CAFASP organizers. From adamz@bialko.llnl.gov Sat Nov 4 01:46:01 2000 Return-Path: Date: Sat, 4 Nov 2000 01:42:33 -0800 (PST) From: adamz@bialko.llnl.gov (Adam ZEMLA) To: karplus@cse.ucsc.edu Subject: CASP4 Abstract Submission Forms Dear Predictor: For the assessors to complete their work and to provide background information for the CASP meeting, we ask that you provide the following details on the prediction methods you used. A. A long abstract (one to two pages), describing the methods you used, and providing any relevant references. You may be able to lift some of the material from the methods information you provided in your prediction submissions. (Please do not rely on the assessors getting the information from that source - in particular cases, the files will be consulted, but generally there are far too many of them!) B. Answers to the specific assessor's questions about your prediction methods. Please fill in the Abstract Submission Form on the web page below by November 10 at the latest. http://PredictionCenter.llnl.gov/casp4/meeting/abstracts.html CASP4 organizers From kaf@sb1.llnl.gov Wed Nov 15 15:48:53 2000 Return-Path: Date: Wed, 15 Nov 2000 15:48:20 -0800 (PST) From: Krzysztof Fidelis To: sam-info@cse.ucsc.edu Subject: Prediction Targets/ Abstracts Dear Predictor: To allow any clarifications regarding your predictions (addressed to the independent assessors) as well as the self-assessment of predictions, we are now making available the coordinates of most of the prediction targets. Coordinates of the remaining targets will be made available as soon as we receive permission from the authors. Please note that most of the structures are still CONFIDENTIAL. Publicly available structures are indicated in the target list available from our website. The structures we are providing now are intended only for the purposes stated above. No further distribution should be made. The structures are available from the following address: http://PredictionCenter.llnl.gov/restricted/Targets_for_predictors Access code: Asilomar2000 Your comments addressed to the assessors will be collected via a special web page. This page will be available beginning Friday, Nov. 17 and will close Tuesday, Nov. 21. I will mail you the address of the page once it is open. Please contact me directly if there is a difficulty with this process. Also, the deadline for submitting methods abstracts as well as poster/software demonstration abstracts has already passed. We have received very many submissions, however, to allow for any late arrivals we will keep this page open until the end of tomorrow, Nov. 16, Pacific Standard Time. Krzysztof Fidelis, For CASP4 Organizers Krzysztof Fidelis, Ph.D. Director, Protein Structure Prediction Center Biology and Biotechnology Research Program Lawrence Livermore National Laboratory L-448, 7000 East Ave., Livermore, CA 94550 Telephone: 925-423-4752 Fax: 925-422-2282 Electronic mail: fidelis@llnl.gov http://PredictionCenter.llnl.gov From kaf@sb1.llnl.gov Fri Nov 17 10:05:15 2000 Return-Path: Date: Fri, 17 Nov 2000 10:04:43 -0800 (PST) From: Krzysztof Fidelis To: karplus@cse.ucsc.edu Subject: Prediction Comments form/ Special CASP session Dear Predictor: The Prediction Comments form is now available. The purpose of this form, as indicated in the preceding announcement, is to allow any clarifications regarding your predictions (addressed to the independent assessors) as well as the assessment of your own predictions. The form is available at the following address: http://PredictionCenter.llnl.gov/casp4/meeting/prediction_comments.html Your comments will be collected until Tuesday, Nov. 21. Please provide additional information if the following applies to your approach. We are planing a special session dedicated to methods derived from first principles, i.e. truly “ab initio” in our CASP sense, and not based on homology of whole proteins or their fragments. Please briefly indicate the approach you have used as well as any results that may be of particular interest to the broader community. You are also welcome to notify us if you consider your approach especially bold, even though the above does not strictly apply. Krzysztof Fidelis, For CASP4 Organizers Krzysztof Fidelis, Ph.D. Director, Protein Structure Prediction Center Biology and Biotechnology Research Program Lawrence Livermore National Laboratory L-448, 7000 East Ave., Livermore, CA 94550 Telephone: 925-423-4752 Fax: 925-422-2282 Electronic mail: fidelis@llnl.gov http://PredictionCenter.llnl.gov From kaf@sb1.llnl.gov Sat Nov 18 19:59:48 2000 Return-Path: Date: Sat, 18 Nov 2000 19:59:16 -0800 (PST) From: Krzysztof Fidelis To: karplus@cse.ucsc.edu Subject: Additional target coordinates Dear Predictor: Targets T0092 (C-alpha coordinates), 97, 105, and 121 have been added to our limited release page at: http://PredictionCenter.llnl.gov/restricted/Targets_for_predictors Access code: Asilomar2000 Please observe structure confidentiality as indicated previously. Krzysztof Fidelis, For CASP4 Organizers Krzysztof Fidelis, Ph.D. Director, Protein Structure Prediction Center Biology and Biotechnology Research Program Lawrence Livermore National Laboratory L-448, 7000 East Ave., Livermore, CA 94550 Telephone: 925-423-4752 Fax: 925-422-2282 Electronic mail: fidelis@llnl.gov http://PredictionCenter.llnl.gov From jmoult@tunc.org Sun Nov 26 05:33:50 2000 Return-Path: X-Sender: jmoult/vipmail.earthlink.net@pop3.norton.antivirus Date: Sun, 26 Nov 2000 08:33:11 -0500 To: karplus@cse.ucsc.edu From: John Moult Subject: Invitation to talk at CASP4 Cc: sippl@came.sbg.ac.at, plo@came.sbg.ac.at Content-Type: text/plain; charset="us-ascii"; format=flowed Dear Kevin: The 'fold recognition' assessor for CASP4 asks that you give a talk in that session. The time allotted will be about 20 minutes, including 5 minutes for questions. As usual at CASP, you will be asked to stick very close to the suggested subjects. A little differently from in the past, each speaker will be introduced by the assessor, clearly indicating the reason that you have been asked to speak, and what he would like to hear you talk about. You will receive instructions directly from the assessor in the next few days. Because of the large number of groups taking part in this CASP and the closeness of many of the results, we have not attempted to select speakers on merit alone, but rather by a mixture of merit and interesting methods. For this reason, please bear in mind that being asked to speak should not be confused with being declared a 'winner' in the way it has in the past. Also, this time, being asked to speak is uncoupled from being asked to write a paper for the special issue. The publication details will be decided on the basis of what goes on at the meeting, and may be a mixture of speaker material and material from the evening discussions. Please let me know as soon as possible whether you agree to speak. I look forward to seeing you at Asilomar. John. From karplus@cse.ucsc.edu Tue Nov 28 14:38:42 2000 Return-Path: Date: Tue, 28 Nov 2000 14:38:30 -0800 From: Kevin Karplus To: jmoult@tunc.org CC: karplus@cse.ucsc.edu Subject: CASP4 talk Of course I agree to speak. I just need some hints what to talk about. Although I've got 35 PDB files from the CASP site, I have no tools for comparing the PDB files with my predictions, and I have no idea which ones I did well on and which ones I did poorly on. Because somewhat different techniques were used on different targets, it would be nice to know which things to concentrate on. Kevin Karplus From jmoult@tunc.org Tue Nov 28 19:57:32 2000 Return-Path: X-Sender: jmoult/vipmail.earthlink.net@pop3.norton.antivirus Date: Tue, 28 Nov 2000 22:56:44 -0500 To: Kevin Karplus From: John Moult Subject: Re: CASP4 talk Cc: sippl@came.sbg.ac.at, plo@came.sbg.ac.at In-Reply-To: <200011282238.OAA10456@purr.cse.ucsc.edu> Content-Type: text/plain; charset="us-ascii"; format=flowed At 02:38 PM 11/28/00 -0800, you wrote: >Of course I agree to speak. >I just need some hints what to talk about. > >Although I've got 35 PDB files from the CASP site, I have no tools for >comparing the PDB files with my predictions, and I have no idea which >ones I did well on and which ones I did poorly on. Because somewhat >different techniques were used on different targets, it would be nice >to know which things to concentrate on. > >Kevin Karplus Kevin: I have passed this to the assessors. John. From karplus@cse.ucsc.edu Wed Nov 29 09:40:03 2000 Return-Path: Date: Wed, 29 Nov 2000 09:39:54 -0800 From: Kevin Karplus To: jmoult@tunc.org, fidelis@llnl.gov CC: karplus@cse.ucsc.edu Subject: post-CASP Symposium Could you send a reminder to all CASP invitees? The UCSC post-CASP symposium is definitely happening. There will be announcments at CASP, but the web page from the original announcment is still there, with up-to-date information: http://www.cse.ucsc.edu/research/compbio/workshop-2000.html From plo@came.sbg.ac.at Wed Nov 29 10:23:16 2000 Return-Path: X-Authentication-Warning: elke.came.sbg.ac.at: plo owned process doing -bs Date: Wed, 29 Nov 2000 19:22:27 +0100 (CET) From: Peter Lackner To: karplus@cse.ucsc.edu cc: John Moult , sippl@came.sbg.ac.at Subject: Re: Invitation to talk at CASP4 In-Reply-To: <5.0.0.25.2.20001126083245.023beeb0@pop3.norton.antivirus> Content-Type: TEXT/PLAIN; charset=US-ASCII Dear Kevin Karplus, Here are the assessment results of your predictions: 95_1 95_2 96_1 100 101 109 110 116_4 121_2 0.5 3 4 3.5 3 1 3 2.5 4 127_1 102 107 108 114 115_1 116_1 116_2 118 3 - 1 0 0.5 0 - 0 1 120_2 126 127_2 87_2 87_1 89_2 90_1 91 94 0 0 2.5 0.5 1 0 - 0 0 96_2 97 104 105 106 115_2 - 0 0 0 - 0 This was done domain based. "_x" in the target id refers to the domain. The model quality is scored in a range of values from 0 to 4. Where 0 is incorrect and 5 would be the ideal model (native). 0 incorrect model 1 model has structural similarity 2 model has correct fold 3 model has correct fold and a medium alignment quality 4 model has correct fold and a good alignment quality Bonus or penalty points (0.5 usually) are given in certain situations, e.g: 1.5 The model has some structural similarities and in this region the alignment is of good quality 0.5 There is some structural similarity, but alignment is poor. We ask you to (1) talk about the good predictions (2) explain what went wrong in the wrong predictions (3) give a short overview of the method(s) you used In addition we ask you to address the following point(s): You have a particularly good prediction for T0127_2, which is a rather difficult target (fold recognition, no homolog in PDB), where the sequence signal should be rather weak. Why does your HMM still find the correct template and how can you get that good alignment. Also, you say that you select models manually. What are the criteria for selection and how to do think this can be automated? Sincerely, the FR assessment team From karplus@cse.ucsc.edu Wed Nov 29 12:52:56 2000 Return-Path: Date: Wed, 29 Nov 2000 12:52:47 -0800 From: Kevin Karplus To: plo@came.sbg.ac.at CC: jmoult@tunc.org, sippl@came.sbg.ac.at, karplus@cse.ucsc.edu In-reply-to: (message from Peter Lackner on Wed, 29 Nov 2000 19:22:27 +0100 (CET)) Subject: Re: Invitation to talk at CASP4 Just confirming---do you really want me to talk about T0127_2? So far as I can see, I did nothing very interesting there. I did a whole-chain search for the target and got 1do0A as the top hit, as did almost all the CAFASP servers. I did move the gaps around by hand, to try to get better conservation, but the alignment of the second domain was within +-2 residues of the automatic aligment T0127-1do0A-2track-local (except at the ends of the domain, where I pulled in some more sequence that the automatic method had rejected). From karplus@cse.ucsc.edu Wed Nov 29 13:18:02 2000 Return-Path: Date: Wed, 29 Nov 2000 13:17:59 -0800 From: Kevin Karplus To: casp4@bialko.llnl.gov CC: sam-info@cse.ucsc.edu In-reply-to: <200011292037.MAA15167@bialko.llnl.gov> (casp4@bialko.llnl.gov) Subject: Re: CASP4 meeting. Secondary structure prediction discussion. I'm willing to make some remarks about what sequences are hard and what sequences are easy to predict, but I don't have a final answer by any means! From casp4@bialko.llnl.gov Wed Nov 29 14:43:20 2000 Return-Path: Date: Wed, 29 Nov 2000 14:39:29 -0800 ("PST) From: casp4@bialko.llnl.gov (CASP4) To: karplus@cse.ucsc.edu Subject: The program for the CASP4 meeting. Dear CASP4 Predictor: The program for the CASP4 meeting has finally been finished. Very tough because of the many high quality predictions. If you have not heard anything, then that means you have not been selected to give a 'morning' talk. There are more prediction groups producing more excellent models than ever before, so this should not in any way be construed as not having 'won'. We will try to decouple speaker choice from that concept as far as possible, and focus the meeting on getting the most out of all these wonderful data. In particular, we are hoping for very lively and informative evening discussions, on topics related to what we need to do to REALLY model proteins structure right. Although we have seen progress over the CASPs, the fact remains that we are still a long, long way from competing with experiment. Come armed with your thoughts about what we can do about that. John Moult, For CASP4 organizers. From plo@came.sbg.ac.at Thu Nov 30 04:02:24 2000 Return-Path: X-Authentication-Warning: elke.came.sbg.ac.at: plo owned process doing -bs Date: Thu, 30 Nov 2000 13:01:35 +0100 (CET) From: Peter Lackner To: Kevin Karplus cc: sippl@came.sbg.ac.at Subject: Re: Invitation to talk at CASP4 In-Reply-To: <200011292052.MAA11281@purr.cse.ucsc.edu> Content-Type: TEXT/PLAIN; charset=US-ASCII On Wed, 29 Nov 2000, Kevin Karplus wrote: Dear Kevin Karplus, > > Just confirming---do you really want me to talk about T0127_2? > So far as I can see, I did nothing very interesting there. > I did a whole-chain search for the target and got 1do0A as the top hit, > as did almost all the CAFASP servers. > > I did move the gaps around by hand, to try to get better conservation, > but the alignment of the second domain was within +-2 residues of the > automatic alignment T0127-1do0A-2track-local (except at the ends of the > domain, where I pulled in some more sequence that the automatic method > had rejected). We didn't suggested to talk about this particular prediction because is a very good one, but still you have got the highest score for this target compared to the other predictors (next best has 2, some few 1.5's). If you prefer to talk about other results, which you think are more interesting, please do so. We really don't want to force you to talk about 127_2. -Peter From karplus@cse.ucsc.edu Thu Nov 30 09:50:31 2000 Return-Path: Date: Thu, 30 Nov 2000 09:50:28 -0800 From: Kevin Karplus To: plo@came.sbg.ac.at CC: sippl@came.sbg.ac.at, karplus@cse.ucsc.edu In-reply-to: (message from Peter Lackner on Thu, 30 Nov 2000 13:01:35 +0100 (CET)) Subject: Re: Invitation to talk at CASP4 I'm perfectly willing to talk about T0127_2, and prepared a couple of slides yesterday to do so. I just wanted to be sure that there hadn't been a mis-communication, since I had thought that 127 was an "easy" target that everyone got. I guess that alignment was tougher than I had thought, so even my partial match was interesting. I was also thinking about discussing T110---not because my score of 3 was impressive (I don't know whether anyone nailed this), but because the method I used for getting it was interesting. Kevin From sippl@came.sbg.ac.at Thu Nov 30 10:04:59 2000 Return-Path: Date: Thu, 30 Nov 2000 19:01:19 +0100 (MET) From: "Manfred J. Sippl" To: Kevin Karplus cc: plo@came.sbg.ac.at Subject: Re: Invitation to talk at CASP4 In-Reply-To: <200011301750.JAA11945@purr.cse.ucsc.edu> Content-Type: TEXT/PLAIN; charset=US-ASCII Dear Kevin: I am currently traveling in the US so communication is difficult. The points you made below are fine. I intend to interview the FR speakers on sunday or monday and go over some issues I would like you to discuss - mostly things that don't need detailed analysis. Looking forward to seeing you in Asilomar, Manfred On Thu, 30 Nov 2000, Kevin Karplus wrote: > > I'm perfectly willing to talk about T0127_2, and prepared a couple of > slides yesterday to do so. I just wanted to be sure that there hadn't > been a mis-communication, since I had thought that 127 was an "easy" > target that everyone got. I guess that alignment was tougher than I > had thought, so even my partial match was interesting. > > I was also thinking about discussing T110---not because my score of 3 > was impressive (I don't know whether anyone nailed this), but because > the method I used for getting it was interesting. > > Kevin > From karplus@cse.ucsc.edu Wed Dec 20 11:48:28 2000 Return-Path: Date: Wed, 20 Dec 2000 11:48:26 -0800 From: Kevin Karplus To: CASP_Complaints@PredictionCenter.llnl.gov CC: karplus@cse.ucsc.edu Subject: Where are CASP results? Other than about 15 minutes at CASP itself, I have been unable to look at the CASP results. Where are the results available? (There doesn't seem to be any link from the CASP4 homepage, except for the CAFASP results.) Where is Manfred's assessment table? That should have been distributed in hardcopy at the conference, but was not, and there is no link to it on the CASP4 site. As a result, I still have no idea how my results compare to anyone else's. I also don't know how the automatic SAM-T99 predictions compare to the hand-done SAM-T2K ones, so I don't know which targets I added value for (and which ones I subtracted value!). This makes it more difficult for me to decide what parts of the hand process to try to automate. From adamz@llnl.gov Wed Dec 20 11:59:10 2000 Return-Path: Sender: adamz@llnl.gov Date: Wed, 20 Dec 2000 11:54:36 -0800 From: Adam ZEMLA Organization: Lawrence Livermore National Laboratory X-Accept-Language: en To: Kevin Karplus Subject: Re: Where are CASP results? Content-Type: text/plain; charset=us-ascii Dear Kevin, there is the link from the main Prediction Center web site: http://PredictionCenter.llnl.gov/Center.html Access: Asilomar2000 We are still waiting for assessors results of evaluation to make it available through the web site. But it looks like before Christmas they will not send us this data. Regards, Adam Kevin Karplus wrote: > > Other than about 15 minutes at CASP itself, I have been unable to look > at the CASP results. > > Where are the results available? (There doesn't seem to be any link > from the CASP4 homepage, except for the CAFASP results.) > > Where is Manfred's assessment table? That should have been > distributed in hardcopy at the conference, but was not, and there is > no link to it on the CASP4 site. As a result, I still have no idea > how my results compare to anyone else's. > > I also don't know how the automatic SAM-T99 predictions compare to the > hand-done SAM-T2K ones, so I don't know which targets I added value > for (and which ones I subtracted value!). This makes it more > difficult for me to decide what parts of the hand process to try to > automate. -- *********************************************************************** Adam Zemla, Ph.D. | E-mail: adamz@llnl.gov Biology and Biotechnology Research Program | Phone: (925) 423-5571 Lawrence Livermore National Laboratory, L-448 | Fax: (925) 422-2282 7000 East Ave., Livermore, CA 94550, USA | Bldg: 3703, Room: 1047 *********************************************************************** From karplus@cse.ucsc.edu Wed Dec 20 13:20:33 2000 Return-Path: Date: Wed, 20 Dec 2000 13:20:31 -0800 From: Kevin Karplus To: CASP_Complaints@PredictionCenter.llnl.gov CC: karplus@cse.ucsc.edu Subject: T0087_2 evaluations scores I just started looking at the CASP4 results (since it was pointed out to me that the link was from http://PredictionCenter.llnl.gov/Center.html and not, as I expected, from the CASp4 web page). I'm a little confused by the T0087_2 evaluations, since according to my notes, I submitted exactly the same alignment for the second domain for both MODEL 1 and MODEL 2, but the evaluation numbers are quite different. The domain boundary was slightly earlier in my predictions than in the assessment, so there shouldn't have been any contamination from the differing first domain. It is quite possible that we made a formatting mistake in the submission, but I'd like someone to look at the submissions and tell us what we did wrong. (Of course, I'd like the correctly evaluated domain 2 to be the evaluation used.) From karplus@cse.ucsc.edu Fri Dec 22 10:48:03 2000 Return-Path: Date: Fri, 22 Dec 2000 10:48:00 -0800 From: Kevin Karplus To: CASP_Complaints@PredictionCenter.llnl.gov CC: karplus@cse.ucsc.edu Subject: Confusion about GDT scores I thought I understood what the GDT scores were, until I looked at the scores for T0116_3, T0116_3A, and T0116_3B. For my models, and for several others, the individual scores for domains 3A and 3B are higher than the score for 3. But since the subdomains are contained in the domain, any superposition that produces n residues with less than A Angstroms for the subdomain will produce >=n residues with less than A angstroms for the whole domain. Thus the doman score should always be at least as high as the subdomain score (though not necessarily as high as the sum of the subdomain scores, since different subdomains may have different optimal superpositions). Can someone explain this anomaly? From casp4@bialko.llnl.gov Fri Dec 22 11:16:22 2000 Return-Path: Date: Fri, 22 Dec 2000 11:11:45 -0800 (PST) From: casp4@bialko.llnl.gov (CASP4) To: karplus@cse.ucsc.edu Subject: Scoring matrix release Dear CASP Participants, the Manfred's CASP4 scoring matrix has been released. It's on the WEB: http://www.came.sbg.ac.at/CASP CASP4 Organizers From casp4@bialko.llnl.gov Fri Dec 22 11:25:15 2000 Return-Path: Date: Fri, 22 Dec 2000 11:20:37 -0800 (PST) From: casp4@bialko.llnl.gov (CASP4) To: karplus@cse.ucsc.edu Subject: CASP4 paper selection still ongoing Dear CASP4 Predictor: This is a follow up to the message of a while ago, concerning the selection of papers for inclusion in the CASP4 special issue. It is now clear this process will not be complete by Christmas, and will continue until early in the New year. John Moult For CASP4 organizers. From adamz@llnl.gov Fri Dec 22 14:54:38 2000 Return-Path: Sender: adamz@llnl.gov Date: Fri, 22 Dec 2000 14:50:01 -0800 From: Adam ZEMLA Organization: Lawrence Livermore National Laboratory X-Accept-Language: en To: Kevin Karplus CC: CASP_Complaints@sb7.llnl.gov Subject: Re: Confusion about GDT scores Content-Type: text/plain; charset=us-ascii Dear Kevin, In GDT_TS calculation the percentage of "well" predicted residues are taken into account, where the target is always used as a frame of reference. If calculations are made on subdomains then subdomain is taken as a frame of reference. GDT scores for the whole target reflect the quality of the whole model. If somebody sent only 70% of the whole structure, but in his model 100% of one of the subdomains was predicted, the scores for the whole target and for the subdomain will differ. The predictor will be somehow penalized for not predicting well the whole structure in the GDT score for the whole target, but because he submitted 100% of the subdomain GDT score for this domain will be higher. Let us check model submitted for target T0116, and the GDT results for subdomains T0116_3, T0116_3a, and T0116_3b: NT - number of residues in target (or subdomain - frame of reference) dom3: 250-542 NT = 293 residues dom3a: 247-385,514-540 NT = 166 residues dom3b: 385-513 NT = 129 residues GDT_Nn - number of residues superimposed under n Angstroms DISTANCE cutoff GDT_TS = 100.0*(GDT_N1 + GDT_N2 + GDT_N4 + GDT_N8)/(4.0*NT) MODEL.TARGET GDT_TS NT NP GDT-N1 GDT-N2 GDT-N4 GDT-N8 T0116AL094_1_1.T0116_3 8.96 293 166 16 17 24 48 T0116AL094_1_1.T0116_3a 15.81 166 106 16 17 24 48 T0116AL094_1_1.T0116_3b 16.09 129 63 7 11 23 42 In GDT analysis each target or subdomain is treated as a separate "target" and all calculations are made on this domain separately. It allows to make proper comparison of all predictions per domain, and per target, and per any other frame of reference. Please let me know if it answers your questions. Best regards, Adam Kevin Karplus wrote: > > I thought I understood what the GDT scores were, until I looked at the > scores for T0116_3, T0116_3A, and T0116_3B. > > For my models, and for several others, the individual scores for > domains 3A and 3B are higher than the score for 3. But since the > subdomains are contained in the domain, any superposition that > produces n residues with less than A Angstroms for the subdomain will > produce >=n residues with less than A angstroms for the whole domain. > Thus the doman score should always be at least as high as the > subdomain score (though not necessarily as high as the sum of the > subdomain scores, since different subdomains may have different > optimal superpositions). > > Can someone explain this anomaly? -- *********************************************************************** Adam Zemla, Ph.D. | E-mail: adamz@llnl.gov Biology and Biotechnology Research Program | Phone: (925) 423-5571 Lawrence Livermore National Laboratory, L-448 | Fax: (925) 422-2282 7000 East Ave., Livermore, CA 94550, USA | Bldg: 3703, Room: 1047 *********************************************************************** From jmoult@tunc.org Tue Jan 30 06:48:46 2001 Return-Path: X-Sender: jmoult/vipmail.earthlink.net@pop3.norton.antivirus Date: Tue, 30 Jan 2001 09:27:32 -0500 To: Kevin Karplus From: John Moult Subject: Invitation to submit a CASP paper. In-Reply-To: <200005161830.LAA16626@purr.cse.ucsc.edu> Content-Type: text/plain; charset="us-ascii"; format=flowed Dear Kevin: I am writing to invite you to submit a paper for consideration for publication in the CASP4 special issue of PROTEINS. The organization of the issue will be similar to that of previous CASPs, with some general papers, and sections on comparative modeling, fold recognition, and new fold methods. Your paper would be in the fold recognition section. As is traditional for the CASP issue, the paper should focus on what went right, what went wrong, and why. In particular, please address as directly and usefully as possible why it is that your results are not perfect. What are the real bottlenecks to progress with your method, and how may they be overcome? (But not too much speculation on the latter!) Because of space restrictions, please do not devote unnecessary space to methods that have already been published. Instead, do link success or failures to particular aspects of your methods. Reviewers will be asked to check that these guidelines have been adhered to. The paper should occupy not more than 6 journal pages. This is a serious space restriction, and reviewers will be asked to ensure that papers conform. To help authors keep within the limit, here are some data on PROTEIN page layout: Column lines are about 8 words wide. There are about 60 lines in a column page. So about 960 words in a full page. Figure captions are in a smaller font: 10 words/line, 9 lines/inch. Don't forget to leave room for the headings and the references. Bear in mind that summary information will be included in the assessors' papers, and so space can be saved in the prediction papers by referring to that. If you are unsure whether relevant information will be present, please consult the relevant assessor. As you know, pictures of the experimental target structures may only be included with the agreement of the experimentalists concerned. Please check the Prediction Center list for current permissions. Please do NOT contact the experimentalists directly on this matter. PROTEINS will cover the cost of a first color page, but additional ones will be charged at $500 a page. The deadline for paper submission is March 1. Because of the production schedule, the deadline should be respected. If this is likely to cause any problems, please let me know now. Papers should be sent directly to the PROTEINS editorial office, and the cover letter should clearly identify them as intended for the CASP4 special issue. Please let me know if you are willing to do this. John Moult, for CASP4 organizers. From karplus@cse.ucsc.edu Tue Jan 30 09:13:29 2001 Return-Path: Date: Tue, 30 Jan 2001 09:13:22 -0800 From: Kevin Karplus To: jmoult@tunc.org CC: karplus@cse.ucsc.edu, rph@cse.ucsc.edu, saira@sanger.ac.uk, cline@cse.ucsc.edu, markd@cse.ucsc.edu, cbarrett@cse.ucsc.edu, stu@cse.ucsc.edu, rachelk@cse.ucsc.edu, jcasper@cse.ucsc.edu In-reply-to: <5.0.0.25.2.20010130091637.025b1e70@pop3.norton.antivirus> (message from John Moult on Tue, 30 Jan 2001 09:27:32 -0500) Subject: Re: Invitation to submit a CASP paper. Yes, I am delighted to be invited to write a paper for the CASP4 special issue. I think that I will concentrate on the differences between SAM-T2K and SAM-T99---why did one do so well relative to the other, since much of the underlying machinery is the same? I've looked at many of the targets where there was a big difference and found a variety of explanations. We are working on automating the most important differences, so that we can have a SAM-T01 server that does as well as the SAM-T2K method did by hand. From sippl@came.sbg.ac.at Wed May 9 04:36:43 2001 Return-Path: Date: Wed, 9 May 2001 13:36:21 +0200 (MEST) From: "Manfred J. Sippl" To: agm@mrc-lmb.cam.ac.uk cc: dbaker@u.washington.edu, m.sternberg@icrf.icnet.uk, sam-info@cse.ucsc.edu, leszek@bioinfo.pl, dfischer@ccs.bgu.ac.il, Burkhard Rost Subject: casp4 FR assessment MS Content-Type: TEXT/PLAIN; charset=US-ASCII Dear all: The CASP FR assessment paper can be downloaded from http://lore.came.sbg.ac.at/CASP/index.html -> assessment paper Best regards, Manfred From nicka@ceres-inc.com Fri May 18 17:55:32 2001 Return-Path: From: nicka@ceres-inc.com To: karplus@cse.ucsc.edu Subject: casp4-cafasp2 confusion Date: Fri, 18 May 2001 17:54:18 -0700 Content-Type: text/plain; charset="iso-8859-1" Dear Kevin, We talked briefly during CASP4 conference about late CAFASP2 predictions which were ignored by CASP4 organizers. My appeal to the CASP4 organizers to include these prediction was not successful. Now I am writing to you in hope that you might have saved email messages from the CASP4 organizers dated by the end of August - middle of September which could clarify this issue. Unfortunately, I have erased this correspondence from my mail box. So far I have found two letters related to this problem: In August, 28, 2000 there was a letter from casp4@bialko.llnl.gov with a subject line: Subject: To all server IDs registered at the prediction center and signed by John Moult, in which they said that CAFASP predictions are not going to be considered by CASP assessors. In September, 14, 2000 there was a letter from Daniel Fischer , in which he said: "- There will be 2 evaluations for cafasp predictions. One carried out by the regular casp assessor..." I remember that there was the third letter, after which I had a clear impression that all my predictions, including CAFASP-late predictions will be assessed. Could you please help me in finding this letter? All I want is just to see my prediction, accepted by CASP4 on the CASP4 evaluation pages. Thank you very much, Nickolai Alexandrov Manager for Bioinformatics Ceres Inc. 3007 Malibu Canyon Road Malibu, CA 90265 e-mail: nicka@ceres-inc.com tel: 310-317-8903 fax: 310-317-8998 This email message is for the sole use of the intended recipient(s) and may contain confidential and privileged information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message. From karplus@cse.ucsc.edu Sat May 19 06:33:05 2001 Return-Path: Date: Sat, 19 May 2001 06:33:00 -0700 From: Kevin Karplus To: nicka@ceres-inc.com CC: karplus@cse.ucsc.edu, casp4@bialko.llnl.gov In-reply-to: (nicka@ceres-inc.com) Subject: Re: casp4-cafasp2 confusion I was not aware that there was still data missing from the CASP assessment. I too am confused by the CASP assessor's refusal to evaluate server predictions that were late for CAFASP, but on time for CASP. There was a clear statement from Dani that late CAFASP predictions would be ignored, but I thought that applied only to the CAFASP evaluations, as he claimed that the two evaluation processes were independent. I looked through my e-mail and could not find any clear statement that the CASP evaluation would also ignore these predictions, nor can I find any rationale for CASP treating the automatic servers differently. One could remove the "automatic server" designation from groups that missed the CAFASP deadline, but removing predictions that made the CASP deadline seems arbitrary and unfair. From adamz@llnl.gov Tue May 22 12:30:57 2001 MIME-Version: 1.0 Sender: adamz@bialko.llnl.gov Date: Tue, 22 May 2001 11:56:43 -0700 From: Adam ZEMLA Organization: Lawrence Livermore National Laboratory X-Accept-Language: en To: Kevin Karplus CC: nicka@ceres-inc.com, adamz@llnl.gov Subject: Re: casp4-cafasp2 confusion Content-Type: text/plain; charset=us-ascii Hi Kevin, Kevin Karplus wrote: > > I was not aware that there was still data missing from the CASP assessment. There is no data missing from CASP assessment. All accepted CASP predictions were evaluated. The owner of ALL predictions submitted for CAFASP experiment (coming from groups who registered as 'servers' at the Livermore site) is CAFASP. All predictions accepted by CAFASP organizers were evaluated by CASP assessors. Predictions submitted for CAFASP experiment were evaluated after CAFASP organizers sent to CASP a list of accepted models. CAFASP organizers have made additional verification of models submitted for CAFASP experiment: - ON TIME requirement (48 hours window to submit RAW data) - equivalence of RAW data and CASP formatted submissions > > I too am confused by the CASP assessor's refusal to evaluate server > predictions that were late for CAFASP, but on time for CASP. > There was a clear statement from Dani that late CAFASP predictions > would be ignored, but I thought that applied only to the CAFASP > evaluations, as he claimed that the two evaluation processes were independent. > > I looked through my e-mail and could not find any clear statement that > the CASP evaluation would also ignore these predictions, nor can I > find any rationale for CASP treating the automatic servers differently. > One could remove the "automatic server" designation from groups that > missed the CAFASP deadline, but removing predictions that made the > CASP deadline seems arbitrary and unfair. Below there is the message that was sent to all predictors during CASP4/CAFASP2 experiment. I hope it explains that these two experiments were independent and the decissions concerning CAFASP models were made by CAFASP organizers. Regards, Adam -------------- Dear Predictor: We are writing to all prediction group leaders who have registered at the Livermore prediction Center as 'server' groups. The purpose of the message is to make very sure that there is no confusion about the relationship between the CAFASP and CASP experiments, and that all prediction submissions are directed appropriately. We believe most of you understand these points fully already, and will not need to take any action. But one case has come to light where there is a misunderstanding, and we want to be sure there are no more. 1. CAFASP and CASP are two separate experiments. Although the results will be compared, a prediction submission received at Livermore is considered to be intended for one or the other, NEVER both. 2. Predictions intended to be part of the CAFASP experiment are taken to be those coming from groups who registered as 'servers' at the Livermore site. These predictions will be compared with the corresponding raw-output stored at the CAFASP site, and if there is a complete match, will be come part of the results of the CAFASP experiment. Again, they will not be considered part of the CASP experiment. 3. All predictions received at the Livermore site that are from non-'server' ID prediction groups will be considered part of the CASP experiment. Again, they will not be considered part of CAFASP. 4. Those predictors wishing to take part in both CASP and CAFASP should have registered as two separate groups at Livermore, one as a 'server', for CAFASP, and one as a regular group, for CASP. The predictions for CAFASP should have been submitted using the CASFASP ID, and the predictions for CASP submitted using the CASP ID. If you have misunderstood any of this, and think you may have sent in predictions that will not be handled appropriately, please do not attempt to re-submit any models. Rather, get in touch with us immediately. We will work with you to try to straighten things out. John Moult For CASP4 organizers. From adamz@llnl.gov Tue May 22 15:02:26 2001 MIME-Version: 1.0 Sender: adamz@bialko.llnl.gov Date: Tue, 22 May 2001 14:55:30 -0700 From: Adam ZEMLA Organization: Lawrence Livermore National Laboratory X-Accept-Language: en To: Kevin Karplus CC: nicka@ceres-inc.com, casp4@bialko.llnl.gov Subject: Re: casp4-cafasp2 confusion Content-Type: text/plain; charset=us-ascii Dear Kevin, thank you for your letter. I want to assure you that no data is missing from CASP, especially from the evaluation summaries. Few days ago I received the following message from Nick Alexandrov > Dear organizers. > I have submitted predictions to CASP4, which were accepted and I could view > these results at your web server in the graphical summary section. However, > I cannot see them now. For example: > > MODEL_INDEX PIN_NUMBER LEADER_CODE ACCEPTANCE DATE > T0086AL389_1 PIN_210419_2418 7344-6288-8179 07/08/00 13:10:49 > > Can you please fix this. > > Regards, > Nickolai Alexandrov First, it is not true that he could view evaluation results for these models. Some of his models were rejected by CAFASP organizers because they didn't meet all CAFASP rules. So, these models were never evaluated. Second, as I know it was explained to him by Dani even during Asilomar meeting why some of his predictions were not accepted by CAFASP organizers. Third, the owner of ALL predictions submitted for CAFASP experiment (coming from groups who registered as 'servers' at the Livermore site) is CAFASP. It was a strong request from CAFASP organizers not to evaluate other models then from their final list of CAFASP accepted models. Finaly, I told Nick that this issue was already discussed and CASP and CAFASP organizers decided that "remaning" CAFASP predictions after evaluation would be posted on the separate WEB page where NO direct comparison with other (CAFASP accepted) predictions will be provided. It will be done after all papers for special PROTEINS issue will be submitted. I hope my letter explains this situation. Regards Adam Kevin Karplus wrote: > > I was not aware that there was still data missing from the CASP assessment. > > I too am confused by the CASP assessor's refusal to evaluate server > predictions that were late for CAFASP, but on time for CASP. > There was a clear statement from Dani that late CAFASP predictions > would be ignored, but I thought that applied only to the CAFASP > evaluations, as he claimed that the two evaluation processes were independent. > > I looked through my e-mail and could not find any clear statement that > the CASP evaluation would also ignore these predictions, nor can I > find any rationale for CASP treating the automatic servers differently. > One could remove the "automatic server" designation from groups that > missed the CAFASP deadline, but removing predictions that made the > CASP deadline seems arbitrary and unfair. -- *********************************************************************** Adam Zemla, Ph.D. | E-mail: adamz@llnl.gov Biology and Biotechnology Research Program | Phone: (925) 423-5571 Lawrence Livermore National Laboratory, L-448 | Fax: (925) 422-2282 7000 East Ave., Livermore, CA 94550, USA | Bldg: 3703, Room: 1047 *********************************************************************** From nicka@ceres-inc.com Tue May 22 17:05:18 2001 MIME-Version: 1.0 From: nicka@ceres-inc.com To: adamz@llnl.gov, karplus@cse.ucsc.edu Cc: nicka@ceres-inc.com Subject: RE: casp4-cafasp2 confusion Date: Tue, 22 May 2001 17:04:05 -0700 Content-Type: text/plain; charset="iso-8859-1" Adam, I understand your arguments, but I am still confused by: 1. I got a acceptance note for disputed predictions from CASP, not CAFASP. 2. The message you attached from CASP organizers was not the only one related to this confusion. Particularly, Dani later had sent a message, where he said (and CASP organizers confirmed their agreement): "- There will be 2 evaluations for cafasp predictions. One carried out by the regular casp assessor, using whatever methods he/she chooses. The other carried out by the cafasp coordinators, using the automated procedures that have been published in the cafasp site." For some (CAFASP-late) predictions the evaluation by CASP assessors have not been done. In contrast to CASP organizers, CAFASP organizers have evaluated late predictions: http://www.cs.bgu.ac.il/~dfischer/CAFASP2/ALevaluation/targetseval.html Nick > -----Original Message----- > From: Adam ZEMLA [mailto:adamz@llnl.gov] > Sent: Tuesday, May 22, 2001 11:57 AM > To: Kevin Karplus > Cc: nicka@ceres-inc.com; adamz@llnl.gov > Subject: Re: casp4-cafasp2 confusion > > > Hi Kevin, > > Kevin Karplus wrote: > > > > I was not aware that there was still data missing from the > CASP assessment. > > There is no data missing from CASP assessment. > All accepted CASP predictions were evaluated. > > The owner of ALL predictions submitted for CAFASP experiment (coming > from > groups who registered as 'servers' at the Livermore site) is CAFASP. > > All predictions accepted by CAFASP organizers were evaluated by CASP > assessors. > > Predictions submitted for CAFASP experiment were evaluated after > CAFASP organizers sent to CASP a list of accepted models. > CAFASP organizers have made additional verification > of models submitted for CAFASP experiment: > - ON TIME requirement (48 hours window to submit RAW data) > - equivalence of RAW data and CASP formatted submissions > > > > > I too am confused by the CASP assessor's refusal to evaluate server > > predictions that were late for CAFASP, but on time for CASP. > > There was a clear statement from Dani that late CAFASP predictions > > would be ignored, but I thought that applied only to the CAFASP > > evaluations, as he claimed that the two evaluation > processes were independent. > > > > I looked through my e-mail and could not find any clear > statement that > > the CASP evaluation would also ignore these predictions, nor can I > > find any rationale for CASP treating the automatic servers > differently. > > One could remove the "automatic server" designation from groups that > > missed the CAFASP deadline, but removing predictions that made the > > CASP deadline seems arbitrary and unfair. > > > Below there is the message that was sent to all predictors during > CASP4/CAFASP2 experiment. I hope it explains that these two > experiments > were independent and the decissions concerning CAFASP models were made > by CAFASP organizers. > > Regards, > Adam > > -------------- > > Dear Predictor: > > We are writing to all prediction group leaders who have registered at > the > Livermore prediction Center as 'server' groups. The purpose of the > message > is to make very sure that there is no confusion about the > relationship > between the CAFASP and CASP experiments, and that all prediction > submissions are directed appropriately. We believe most of you > understand > these points fully already, and will not need to take any action. But > one > case has come to light where there is a misunderstanding, and > we want to > be > sure there are no more. > > 1. CAFASP and CASP are two separate experiments. Although the results > will > be compared, a prediction submission received at Livermore is > considered > to > be intended for one or the other, NEVER both. > > 2. Predictions intended to be part of the CAFASP experiment > are taken to > be > those coming from groups who registered as 'servers' at the Livermore > site. > These predictions will be compared with the corresponding raw-output > stored > at the CAFASP site, and if there is a complete match, will > be come part > of > the results of the CAFASP experiment. Again, they will not be > considered > part of the CASP experiment. > > 3. All predictions received at the Livermore site that are from > non-'server' ID prediction groups will be considered part of the CASP > experiment. Again, they will not be considered part of CAFASP. > > 4. Those predictors wishing to take part in both CASP and > CAFASP should > have registered as two separate groups at Livermore, one as a > 'server', > for > CAFASP, and one as a regular group, for CASP. The predictions > for CAFASP > should have been submitted using the CASFASP ID, and the > predictions for > CASP submitted using the CASP ID. > > If you have misunderstood any of this, and think you may have sent in > predictions that will not be handled appropriately, please do not > attempt > to re-submit any models. Rather, get in touch with us immediately. We > will > work with you to try to straighten things out. > > John Moult > For CASP4 organizers. > From adamz@llnl.gov Tue May 22 18:49:18 2001 MIME-Version: 1.0 Sender: adamz@bialko.llnl.gov Date: Tue, 22 May 2001 18:42:23 -0700 From: Adam ZEMLA Organization: Lawrence Livermore National Laboratory X-Accept-Language: en To: nicka@ceres-inc.com CC: karplus@cse.ucsc.edu, adamz@llnl.gov Subject: Re: casp4-cafasp2 confusion Content-Type: text/plain; charset=us-ascii nicka@ceres-inc.com wrote: > > Adam, I understand your arguments, but I am still confused by: > > 1. I got a acceptance note for disputed predictions from CASP, not CAFASP. CAFASP asked us to collect for them formatted submissions. They were not ready that time to provide service for collecting formatted predictions. But they reserved all rights to predictions submitted from groups registered as 'servers'. > > 2. The message you attached from CASP organizers was not the only one > related to this confusion. Particularly, Dani later had sent a message, > where he said (and CASP organizers confirmed their agreement): > > "- There will be 2 evaluations for cafasp predictions. One carried out by > the regular casp assessor, using whatever methods he/she chooses. The other > carried out by the cafasp coordinators, using the automated procedures > that have been published in the cafasp site." Yes, but it was a strong request from CAFASP organizers not to evaluate other models than from their own final list of CAFASP accepted models. And some of your models were not on the list provided by CAFASP organizers. > > For some (CAFASP-late) predictions the evaluation by CASP assessors have not > been done. None of the CAFASP-late predictions were evaluated by CASP assessors. On the CASP site the only CASP models and models accepted by CAFASP were evaluated. > In contrast to CASP organizers, CAFASP organizers have evaluated > late predictions: > http://www.cs.bgu.ac.il/~dfischer/CAFASP2/ALevaluation/targetseval.html As you know very well CAFASP organizers didn't provide this data for CASP assessors, and results on CAFASP web site appeared probably just after our request to allow us to evaluate LATE CAFASP predictions. As I told you such agreement was finally made recently and evaluation results using CASP evaluation criteria will be provided on our site soon. Adam PS. One more question: Nick, why haven't you discussed this subject so strongly earlier with CAFASP organizers? You knew very well during Asilomar meeting that it was CAFASP decision not to provide CASP and CASP assessors with LATE models. I wonder why you are making CASP responsible for CAFASP rules and their decisions. From nicka@ceres-inc.com Wed Jun 6 18:15:38 2001 MIME-Version: 1.0 From: nicka@ceres-inc.com To: adamz@llnl.gov, nicka@ceres-inc.com Cc: karplus@cse.ucsc.edu Subject: RE: casp4-cafasp2 confusion Date: Wed, 6 Jun 2001 18:14:34 -0700 Content-Type: text/plain; charset="iso-8859-1" Adam, during our telephone conversation on May 23, you agreed to put missing predictions at the CASP4 evaluation site within two or three weeks. I hope you did not change your mind and I will see my prediction evaluated soon. Nick > -----Original Message----- > From: Adam ZEMLA [mailto:adamz@llnl.gov] > Sent: Tuesday, May 22, 2001 6:42 PM > To: nicka@ceres-inc.com > Cc: karplus@cse.ucsc.edu; adamz@llnl.gov > Subject: Re: casp4-cafasp2 confusion > > > nicka@ceres-inc.com wrote: > > > > Adam, I understand your arguments, but I am still confused by: > > > > 1. I got a acceptance note for disputed predictions from > CASP, not CAFASP. > > CAFASP asked us to collect for them formatted submissions. They were > not ready that time to provide service for collecting formatted > predictions. But they reserved all rights to predictions > submitted from > groups registered as 'servers'. > > > > > 2. The message you attached from CASP organizers was not > the only one > > related to this confusion. Particularly, Dani later had > sent a message, > > where he said (and CASP organizers confirmed their agreement): > > > > "- There will be 2 evaluations for cafasp predictions. One > carried out by > > the regular casp assessor, using whatever methods he/she > chooses. The other > > carried out by the cafasp coordinators, using the automated > procedures > > that have been published in the cafasp site." > > Yes, but it was a strong request from CAFASP organizers not > to evaluate > other models than from their own final list of CAFASP > accepted models. > And some of your models were not on the list provided by CAFASP > organizers. > > > > > For some (CAFASP-late) predictions the evaluation by CASP > assessors have not > > been done. > > None of the CAFASP-late predictions were evaluated by CASP assessors. > On the CASP site the only CASP models and models accepted by CAFASP > were evaluated. > > > In contrast to CASP organizers, CAFASP organizers have evaluated > > late predictions: > > http://www.cs.bgu.ac.il/~dfischer/CAFASP2/ALevaluation/targetseval.html As you know very well CAFASP organizers didn't provide this data for CASP assessors, and results on CAFASP web site appeared probably just after our request to allow us to evaluate LATE CAFASP predictions. As I told you such agreement was finally made recently and evaluation results using CASP evaluation criteria will be provided on our site soon. Adam PS. One more question: Nick, why haven't you discussed this subject so strongly earlier with CAFASP organizers? You knew very well during Asilomar meeting that it was CAFASP decision not to provide CASP and CASP assessors with LATE models. I wonder why you are making CASP responsible for CAFASP rules and their decisions.