Wed Jul 19 09:19:57 PDT 2006 T0386 Make started Wed Jul 19 09:20:43 PDT 2006 Running on lopez.cse.ucsc.edu Wed Jul 19 10:12:13 PDT 2006 Kevin Karplus Both BLAST and the HMMs are getting 2g03A and 2f6sA as strong hits. (BLAST 2g03A 25% over 206 residues, E-value 8e-04). Wed Jul 19 16:06:47 PDT 2006 Kevin Karplus This seems to be 7 helices packed around a central helix. I hope that undertaker doesn't blow it up. Thu Jul 20 10:39:18 PDT 2006 Kevin Karplus We seem to be aligning about 5 of the 7 helices (D64-Q211). Wed Aug 2 14:05:44 PDT 2006 Zack Sanborn Well, it's about time to get started on this protein, seeing as it is not only Grant's and my last target, but also the LAST target of CASP! Well, looking at the PDB-BLAST hit 2g03A: ------------------------------------------------------- Leisseria Meningitidis. Length=194 Score = 38.1 bits (87), Expect = 0.002 Identities = 43/164 (26%), Positives = 70/164 (42%), Gaps = 32/164 (19%) Query 65 TAYLCHIHQQLFKNTFEWAGYLRHIPFTFADGTTAAMPEMKRTGWKNAFAIGDEIQEGLQ 124 TA L IH+ LF +++AG +R + + G++ FA ++E L Sbjct 42 TAGLQQIHRYLFGGLYDFAGQIRE-------------DNISKGGFR--FANAXYLKEALV 86 Query 125 RLDQTLAEKNNLQGLTREEFNSEAIELFNSLNQLHPFREGNGRTQRLFFE-NLAKAAGHQ 183 +++Q E E I + N HPF EGNGR+ R++ + L K Sbjct 87 KIEQX-----------PERTFEEIIAKYVEXNIAHPFLEGNGRSTRIWLDLVLKKNLKKV 135 Query 184 LNFSLITKERMMVASVAVAENGDLE----PMQHLFEDISNPEKI 223 +N+ ++K + A N DLE +L +D+ N E I Sbjct 136 VNWQNVSKTLYLQAXERSPVN-DLELRFLLKDNLTDDVDNREII 178 ------------------------------------------------------- We see that 2g03 matches the center of our protein, but has about 50-75 residues on either side that don't align. Looking at the try1-opt2 model, we see that there are some helices that just don't where to be placed. Grant had an idea to make a set of three overlapping "subdomains." We understand that this protein isn't actually multidomain, but we hope that the prediction for the middle part will be really good and we can use the predictions for the surrounding domains to make a chimera of the full protein. Of course, it may not work. The middle domain is found T65-I223. So, the overlapping domains I chose are: M1 - L75 First "domain" A50 - E250 Second "domain" matching 2g03 with buffer L225 - T299 Third "domain" I've made the subdomains, and started the intial predictions using the command: ( make -k >& make.log; gzip -9f make.log ) T0386/M1-L75 started on orcas. T0386/A50-E250 started on whidbey. T0386/L225-T299 started on whidbey. Fri Aug 4 11:18:37 PDT 2006 Zack Sanborn Well, making the subdomains wasn't the cure-all that we wanted it to be. The middle subdomain (A50-E250) still looks pretty bad. The other subdomains look better, but we're not terrible confident in the predictions. Before we can start putting the full model together, we need to fix that middle subdomain. It's a helix bundle that has been splayed out by Undertaker. We need to put in some distance constraints to keep those helices together. I'm going to do this using our subdomain prediction on A50-E250, since it has the fewest residues to deal with. I'm using Rasmol, picking the distance constraints from model 1 of T0386.undertaker-align.pdb (the top template hit 2g03). I came up with the following distance constraints, now I'll format them properly for Undertaker: ALA180A.CB-LEU139A.CB: 6.499 PHE173A.CB-PHE152A.CB: 4.898 GLN169A.CB-LEU155A.CB: 5.633 THR129A.CB-GLU147A.CB: 4.231 LEU126A.CB-SER154A.CB: 4.777 ASN145A.CB-GLN211A.CB: 5.525 GLU215A.CB-ARG141A.CB: 6.475 PRO209A.CB-ALA197A.CB: 4.145 LEU213A.CB-ARG193A.CB: 5.435 ARG141A.CB-GLU215A.CB: 6.475 LEU213A.CB-ARG193A.CB: 5.435 ARG167A.CB-PHE229A.CB: 5.583 MET51A.CB-LEU226A.CB: 5.145 LYS178A.CB-LYS222A.CB: 11.264 LEU75A.CB-ARG167A.CB: 6.203 ARG170A.CB-PHE229A.CB: 5.141 ASN145A.CB-GLN211A.CB: 5.525 ASP127A.CB-THR65A.CB: 4.787 ALA180A.CB-PHE144A.CB: 6.388 SER218A.CB-ALA177A.CB: 7.326 GLN211A.CB-ASN145A.CB: 5.525 LEU158A.CB-LEU126A.CB: 5.941 ASN175A.CB-TYR67A.CB: 7.251 GLN169A.CB-LEU155A.CB: 5.633 Fri Aug 4 11:34:04 PDT 2006 Grant Thiltgen I'm looking at the first region. We're getting some really weak strand predictions for the first region, which I suppose are better than nothing. I do like the first model from undertaker, so it's probably good for making a chimera out of, but I will retry with some constraints for the strands and make a sheet. There are also a few hbond predictions. Fri Aug 4 12:07:24 PDT 2006 Grant Thiltgen Try2 is set up for the first domain M1-L75. I added in Helix and strand constraints from the secondary structure predictions, and a few hbond constraints. Try2 started on whidbey. Fri Aug 4 12:08:57 PDT 2006 Zack Sanborn I've made the distance constraints and placed them in try2.costfcn along with the try1-opt2 helix and sheet constraints. This run will restart from the alignments, hopefully it works: A50-E250/try2 started on orcas. Fri Aug 4 12:45:00 PDT 2006 Grant Thiltgen For the third domain, there aren't any good pdb hits or any good fold recognition hits either. I may have to take an ab initio approach to this one too. The original model made a barrel, which may or may not be right. The secondary structure predictions for this region do create a bunch of anti-parallel strands so it could be right. The overlapping region in this model does create a helix, so that's a good sign. We're getting some decent sep predictions for the end of the strands, so this could help us create the strands correctly. Checking the predictions, they actually line up perfectly with the strands. There is one strand that's predicted to be a helix, so I need to check into that one and see why it's getting that prediction. It actually looks really good for that region. I'll start another try on it to see what we get. Also, if the middle region has the beginning and the end of the domain in the same area, the two domains could interact, but I'll check into that as soon as the second try for the second domain finishes. There are some strong RR predictions, and there are two red strands with the rr script. There is also a strongly predicted residue in another strand predicted too, which is on the opposite side of the sheet, so I am going to try to get the strands to form the sheet on the opposite side. Fri Aug 4 13:52:50 PDT 2006 Grant Thiltgen try2 for the third domain is finished. I put in constraints to put the two strands on the other side of the sheet. I also added some hbond constraints to keep things together. Try2 started on whidbey. Fri Aug 4 14:00:08 PDT 2006 Zack Sanborn Well, while the large, middle subdomain is still running, I'll take a look at the first domain. try2 has finished, and it did not make the strands that Grant wanted. Why don't I give it a shot! Okay, that was easy. Looking at try2.costfcn, we realized that Grant put in "StrandConstraint"s instead of "SheetConstraint"s. So, I've fixed that in try3.costfcn, and copied try2 to try3 to make an identical run. Also, hedging our bets, I increased the constraints to 20. M1-L75/try3 started on camano. Fri Aug 4 14:45:43 PDT 2006 Grant Thiltgen A225-T299 try2 didn't work so well, the sheets didn't move. I am increasing the weight of the strands I want to move to the other side of the sheet in the costfcn and I'm re-running it. I also put in the wrong residue for the constraints, which could be why it didn't work. Try3 started on whidbey. Fri Aug 4 15:18:40 PDT 2006 Zack Sanborn Try3 on the M1-L75 subdomain has finished and managed to make the sheets we wanted. Since the middle domain is still running, I'm going to do another run, starting from try3-opt2.gromacs0 and try to clean up the breaks. M1-L75/try4 started on shaw. Fri Aug 4 15:38:48 PDT 2006 Zack Sanborn try2 on the middle subdomain has finished and looks very good! We superimposed try2-opt2 with model 1 of T0386.undertaker-align.pdb (the model where we took the distance constraints) and they overlap VERY well... surprisingly well, in fact. However, the model has a few breaks that we'd like to clean up before moving on to making a full model. A50-E250/try3, starting with try2-opt2 and upping breaks and soft_clashes, started on orcas. Fri Aug 4 16:18:42 PDT 2006 Zack Sanborn Oops, just realized that try4 was all screwed up (forgot to globally replace try3 with try4, I almost NEVER do that). I've fixed it and restarted M1-L75/try4 on shaw. Fri Aug 4 17:18:19 PDT 2006 Grant Thiltgen Try3 on the third domain looks pretty good, but the sheet needs to form a bit better. I'm restarting with try4 to see if I get the same results or something different. try4 started on whidbey. Fri Aug 4 18:17:45 PDT 2006 Zack Sanborn Well, M1-L75/try4 and A50-E250/try3 have finished. A50-E250/try3-opt2 does VERY well on breaks, with the largest one less than 0.5 I'm not sure if we want to do any better than this before putting the full model together. So, I've started a try2 for the full protein. This try2 will restart from the alignments but using the sheet, helix, and residue constraints from each of the subdomains. It is hoped that this will produce a good model. We may also make a chimera of these three subdomains. This may prove difficult since the N- and C-terminal ends of the middle subdomain are in roughly the same location. This could mean that the outer subdomains interact with eachother. The above try2 might make this moot as it will hopefully find a way that these subdomains might interact. We'll see. try2 started on orcas. Sat Aug 5 02:42:08 PDT 2006 Kevin Karplus What did the servers do with this target? How do the server models score with the try2 costfcn? I'll make decoys/score-all+servers.try2.pretty Sat Aug 5 03:23:41 PDT 2006 Kevin Karplus Several of the Robetta models, and even SAM_T06_server_TS1 score much better with try2.costfcn than any of the models we've built so far. Sat Aug 5 12:18:07 PDT 2006 Zack Sanborn Well, it looks like Undertaker has struck again! It has placed sheets in our nice helix bundle. I think I need to increase all of the constraints on the middle subdomain relative to the constraints on the outer subdomains and see where that takes me. While I'm at it, I'll include the helix and sheet constraints from the latest L225-E229/try4-opt2 subdomain model. try3, on the full protein, started on shaw. Sat Aug 5 12:35:03 PDT 2006 Zack Sanborn Well, I checked out the top 2 server models (SAM_T06_server_TS1 and ROBETTA_TS3) using the try2.costfcn. Incidentally, SAM_T006_server_TS1 is also the highest scoring model with an unconstrained cost function. I superimposed these models against try1-opt2, try2-opt2, and model 1 of T0386.undertaker-align.pdb (alignment to 2g03). Both of the server models exhibit our desired helix bundle in the middle subdomain. If try2-opt2 had worked, and kept that subdomain from A50-E250 intact, then I'm pretty sure try2-opt2 would have compared favorably to these automatic server predictions. I like our server prediction SAM_T06_server_TS1 better than the ROBETTA_TS3 model simply because the ROBETTA model has a lot of loose coils dangling all over the place. I think it might help to do a polishing run of SAM_T06_server_TS1, cleaning up breaks, soft_clashes, improving packing, etc. try4, optimizing SAM_T06_server_TS1 for breaks, clashes, etc., started on orcas. Sat Aug 5 13:36:54 PDT 2006 Zack Sanborn I've made a terrible looking chimera by attempting to superpose the three subdomains, using the script make-chimera.under. It put the superposition in chimera-models.pdb. However, it looks like Undertaker can't superimpose partial models. But, Undertaker got close enough for the 1st and 2nd domains to make a decent chimera. The third domain was placed far away from the rest of the model. I'm hoping Undertaker will be able to reduce the huge gap caused by this separation and find a good placement for this domain. The chimera model is found in decoys/T0386.chimera.pdb. try5, optimizing the chimera, was started on shaw. I've retained the constraints from all of the original subdomain models in try5.costfcn to keep Undertaker from destroying all of our nice work. Sat Aug 5 16:59:23 PDT 2006 Grant Thiltgen I really dislike try4, which is the model we built the first chimera from. I am rerunning the subdirectory with try5, to see if I can get the sheets made right. Try1 in L225-T299 still looks really good, so we can use that too. Sat Aug 5 17:05:16 PDT 2006 Zack Sanborn I've made a 2nd chimera of the three subdomains. Grant told me what I was doing wrong with my initial stab at superposing the three subdomains. Apparently, you can't superpose two different locations unless all three models share those residues. My second chimera was built from M1-D64/try4-opt2, T65-R224/try3-opt2, and L225-T299/try1-opt2. Our initial attempt at this chimera used L22-T299/try4-opt2, but it was decided (see above) that it was not a terribly good model. The chimera can be found in decoys/T0386.chimera2.pdb. I've started a try6 that optimizes this chimera on shaw. Sat Aug 5 17:18:05 PDT 2006 Zack Sanborn Well, since Grant went to all the trouble of fixing our troubled chimera using L225-T299/try4-opt2 using ProteinShop to move the third domain OUT of the middle of the second domain, we may as well optimize it for model variety's sake. This chimera was made from M1-D64/try4-opt2, T65-R224/try3-opt2, and L225-T299/try4-opt2. It can be found in decoys/T0386.chimera3.pdb. try7, optimizing T0386.chimera3.pdb, was started on orcas. Sat Aug 5 21:41:13 PDT 2006 Kevin Karplus Grabbing some constraints from align1 I52.CB L226.CB 7.6 A48.CB M230.CB 7.2 A48.CB L171.CB 7.5 I71.CB L171.CB 4.9 L68.CB F172.CB 5.0 L155.CB Q169.CB 5.6 F152.CB F173.CB 4.9 F214.CB F173.CB 8.5 N153.CB L207.CB 5.3 M210.CB N145.CB 6.1 F214.CB L184.CB 5.6 F229.CB L171.CB 7.7 Q169.CB N156.CB 6.1 Sat Aug 5 22:00:34 PDT 2006 Kevin Karplus I started an optimization from the alignments with those distances constrained and the dssp-ehl2 constraints (no sheets and no rr constraints). I included the alignments from the "subdomains" also. try8 started on cheep. Sun Aug 6 03:20:57 PDT 2006 Kevin Karplus try8-opt2 does not score particularly well, even with the try8 costfcn, which prefers try4-opt2. Rosetta also likes best try4-opt2.gromacs0.repack-nonPC, so try4 may be our best. I'd better look at it. try4 does retain the core of the alignments pretty well. It comes from polishing our SAM_T06_server_TS1 model, so it is really not cheating. (I was worried that it came from a robetta model, as those seem to retain the core also---probably becasue they were done with a frozen core). Sun Aug 6 04:11:15 PDT 2006 Kevin Karplus The conserved motif responsible for most of the fold recognition seems to start at H159: HPFx[DE]GNGRxxR, so I'll superimpose the models based on that conserved motif. All the models currently in superimpose-best.under have at least reasonable similarity to the templates in the area around the conserved motif, though my try8-opt2 is probably the worst. I picked the order of the "try models" based on the score on the try8 costfcn (which incidentally likes the same two server models that had previously been identified. ReadConformPDB T0386.try4-opt2.gromacs0.repack-nonPC.pdb ReadConformPDB T0386.try6-opt2.pdb ReadConformPDB T0386.try3-opt2.pdb ReadConformPDB T0386.try8-opt2.pdb ReadConformPDB T0386.try1-opt2.pdb ReadConformPDB T0386.try7-opt2.pdb InFilePrefix decoys/servers/ ReadConformPDB SAM_T06_server_TS1.pdb.gz ReadConformPDB ROBETTA_TS3.pdb.gz It looks like we have fold recognition for the central bundle of helices, but new-fold prediction for the rest. It may be worth making a chimera of try4-opt2 and something for T236-T299 (perhaps taken from the L225-T299 subdirectory). The N-terminus may be messed up in try4-opt2, but not nearly as badly as the C-terminus. I have to grab some breakfast and go to my meeting, but I may be able to work on this before others get up. Sun Aug 6 10:26:13 PDT 2006 Kevin Karplus I made make-chimera2.under to superimpose try4-opt2.gromacs0.repack-nonPC and L225-T229/try1-opt2 to make for-chimera-4-1.pdb I'll do the crossover between F229 and M230. Sun Aug 6 10:42:33 PDT 2006 Kevin Karplus try9 to optimize (with constrained optimization) chimera-4-1 started on cheep. Sun Aug 6 11:06:54 PDT 2006 Kevin Karplus I just realized that for try9 I included a lot of alignments, but I did not include the fragments from the subdomains. When I do the unconstrained polishing run (try10, after try9 finishes), I will include all the fragments, but not the alignments. Sun Aug 6 11:50:01 PDT 2006 Grant Thiltgen I thought I started try5 for L225-T299 last night, but I guess I missed it. I started try5 for the subdomain on vashon. I'm also starting to get constraints for the second alignments for the middle subdomain, so try4 will be working with the second alignments, like getting distance constraints from them. Sun Aug 6 12:14:03 PDT 2006 Zack Sanborn I've made a NEW chimera from M1-L75/try1-opt2 (all helix, as opposed to a helix/beta model of try3-opt2), and the same 2nd and 3rd subdomains as try6-opt2 (A50-E250/try3-opt2 & L225-T299/try1-opt2). I picked the crossover point for the domains at T64/A65 and L224/L225. The chimera is currently stored in decoys/ProteinShop/chimera-model2-stage2.pdb. Unfortunately, the reason why it's sitting in the ProteinShop directory and not the decoys directory is that the first domain needs to be ProteinShop'd out of the rest of the protein. It is currently sitting in the middle of the second subdomain. Grant will ProteinShop it once he's done with the work he described above. I've set up a try11 to do the optimization of this chimera. It should be all ready to go. While we wait for Grant, and since we have a LOT of computer's at our disposal I'm going to start a couple polishing runs of our current favorite models, so we have some nice ones to submit. I'm going to pick Undertaker's favorite model with an unconstrained cost function (try6-opt2) and Rosetta's favorite model (try4-opt2.gromacs0.repack-nonPC) for the optimization. try12, optimizing try6-opt2 for breaks, soft_clashes, etc. started on orcas. try13, optimizing try4-opt2.gromacs0.repack-nonPC, started on lopez. Sun Aug 6 12:26:04 PDT 2006 Grant Thiltgen A50-L225/try4 is finished, and has constraints for the helices from the 2nd alignment. The third alignment looks very different in comparison to the first two, and since we are getting such a high PDB hit, I don't think doing models that are too different in the core region is a good idea. Try4 started on shaw. Sun Aug 6 12:36:25 PDT 2006 Zack Sanborn Grant finished with fixing the chimera in the file chimera-model2-stage2-proteinshop.pdb. I've moved it to decoys/T0386.chimera4.pdb.gz. I've started try11, optimizing the new chimera, on lopez. Sun Aug 6 12:43:10 PDT 2006 Grant Thiltgen Try5 in the L225-A299 subdomain looks reasonable to make into a chimera. I still don't like the end that I'm trying to fix the sheets on, I may add some residue constraints to try to move them closer. Try6 in the L225-A299 added some residue constraints from the rr predictions to see if the sheet forms better. Try6 started on vashon. Sun Aug 6 13:46:24 PDT 2006 Kevin Karplus try9-opt2 scores best with the try9 and try10 cost fcns. Note that try10.costfcn has no constraints, and has a high weight for breaks and clashes, so it is a fairly appropriate "neutral" costfcn for evaluating final models. try9-opt2 still has some bad breaks: T0386.try9-opt2.pdb.gz has 39 breaks T0386.try9-opt2.pdb.gz breaks before (T0386)A100 with cost 4.39145 T0386.try9-opt2.pdb.gz breaks before (T0386)M101 with cost 3.73415 T0386.try9-opt2.pdb.gz breaks before (T0386)T255 with cost 1.74799 T0386.try9-opt2.pdb.gz breaks before (T0386)K222 with cost 1.58803 T0386.try9-opt2.pdb.gz breaks before (T0386)E221 with cost 1.09825 T0386.try9-opt2.pdb.gz breaks before (T0386)A99 with cost 0.915632 T0386.try9-opt2.pdb.gz breaks before (T0386)P60 with cost 0.80293 T0386.try9-opt2.pdb.gz breaks before (T0386)D294 with cost 0.737939 T0386.try9-opt2.pdb.gz breaks before (T0386)N23 with cost 0.731451 T0386.try9-opt2.pdb.gz breaks before (T0386)L287 with cost 0.634324 So try10 will try to polish try9-op2.gromacs0 and try9-opt2.gromacs0.repack-nonPC. try10 started on cheep. Because try9 has a good score with try10.costfcn already, I'm fairly sure that the try10 model will not blow up even optimizing with no constraints. Sun Aug 6 14:05:05 PDT 2006 Kevin Karplus One part of try9-opt2 that I'm *not* impressed with is L225-M230, which has unwound the helix a bit. Sun Aug 6 16:14:15 PDT 2006 Zack Sanborn We made some more chimeras! Using the make-chimera4.under script: T0386.chimera4.pdb was made from M1-T64/try4-opt2, A65-L224/try4-opt2, and L225-T299/try5-opt2. T0386.chimera4.pdb was made from M1-T64/try4-opt2, A65-L224/try4-opt2, and L225-T299/try6-opt2. T0386.chimera4.pdb was made from M1-T64/try4-opt2, A65-L224/try4-opt2, and L225-T299/try1-opt2. try14, optimizing chimera4, started on camano. try15, optimizing chimera5, started on camano. try16, optimizing chimera6, started on orcas. All of the above trys have had their OptConform's reduced and are reading in fragments from all of the subdomains in an effort to get models more quickly. These try.under scripts and .costfcns were modeled (read: copied) from try10.under. Sun Aug 6 19:55:54 PDT 2006 Grant Thiltgen Now that the chimeras have run, I am going to run a quick polishing run on them to see if they get any better. Try17 will polish try14-opt2 Try18 will polish try15-opt2 Try19 will polish try16-opt2. Since the main core is comparative modeling, I'm sure we have the main section right, but with the small regions not picking up anything decent, I hope the ab initio approach works. Try17 and try18 started on shaw. Try19 started on lopez. Damn. Undertaker tore apart the first subdomain of the fifth and sixth chimera. It looks like Zack may have not used try4-opt2 for the first subdomain in chimera 4 though. I am going to have to re-polish those models, but I am going to have to include the constraints to keep the regions intact I think. I'm not going to bother with chimera4, but chimera5 and chimera6 I am going to try to re-run through undertaker. Try20 which is a polish of chimera5 is started on lopez. Try21 which is a polish of chimera6 is started on whidbey. Sun Aug 6 22:35:47 PDT 2006 Grant Thiltgen All the runs are finished. I have decided on these five models for the top models: try10-opt2 try12-opt2 try13-opt2 try20-opt2 try17-opt2.gromacs0.repack-nonPC To pick this order, I used a number of different costfcns. Try10.costfcn, unconstrained.costfcn, and the rosetta scoring function. Try10-opt2 scored best on the try10 costfcn (not that that's a huge surprise) and try10-opt2.gromacs0.repack-nonPC scores best with the rosetta scoring function, so I decided to place that one first. Try12-opt2 scores second best in the try10.costfcn and scores best with the unconstrained costfcn, so I went with this one second. Try13-opt2 is the third highest scoring model in the try10 costfcn, and scores decent with the unconstrained costfcn Try20-opt2 is the second best scoring model is the unconstrained costfcn (Behind try6 which is similar to try12). The last model I included was try17-opt2. This model is a polish of try14, which was the second highest scoring model in the rosetta scoring function. Try17 scores better than try14 in the try10 costfcn and scores slightly worse than try14 in the unconstrained costfcn, so I opted to include one of the two here. Since try17 scores better in try10, I went with this one, and I included the rosetta scoring one because that's how I chose this model. ------------ For this model, we had two strong hits to the PDB and to our HMM's, 2g03A and 2f6sA. The best hit PDB hit to 2g03A matched residues 65-223 of the target. However, since the target is 299 residues, we needed to split the model into domains for the first region (residues 1-65) and the last region (residues 223-299). These two regions had no close PDB hits and poor fold recognition hits, so we had to model these using ab initio techniques. For almost all the models we made chimeras of the three domains in order to make decent models, and we had to include constraints to keep the seven helix bundle for the comparative modelling portion of the protein together. Model 1 is try10-opt2. This is a chimera of try4, which was the first polish of the best SAM_T06 server model, and using the first run of our subdomain from L225-T299. We included residues 1-229 of the polished SAM_T06 model and residues 230-299 of the subdomain we ran. This model scores the best using the rosetta scoring function. It also scores best using the try10 costfcn, which is a fairly neutral function that we decided to use. Model 2 is try12-opt2. This model is also built from a chimera. The chimera was built by using three separate predictions from subdomains. The first subdomain was from M1-D64, and used a model based on secondary structure prediction and hydrogen bond predictions in the first subdomain. Residues T65-R224 used try3-opt2, which was a polished model built from alignments and using distance constraints to keep the seven helix bundle together. The alignments and constraints were from 2g03A. The third domain was from L225-T299 and used try1-opt2 from this region, which was the original results from our first run in this subdomain. It scored well on two costfcns that we used to analyze the models. It scored first on the unconstrained costfcn and scored second best on the try10 costfcn. Model 3 is try13-opt2. This model is a polish of the complete best scoring server model from the SAM_T06 server. It kept the seven helix bundle in the center of the protein and model the rest of the protein fairly nicely. Model 4 is try20-opt2. This model is a polish of another chimera that we had built. This chimera was also built from three subdomains. M1-T64 used try4-opt2, which was the model based on secondary structure predictions. A65-L224 used try4-opt2 which was another model built with constraints on helices and alignments, but this time using constraints from the other well scoring model, 2f6sA. L225-T299 used try6-opt2, which attempted to use predicted residue-residue contacts, predicted hydrogen bonds, and predicted secondary structure. This chimera was optimized and polished in undertaker. This model scored second best on the unconstrained costfcn and scored decently on the try10 costfcn. Model 5 is try17-opt2.gromacs0.repack-nonPC. This model scores very well with rosetta. I included it here for a bit of variety for submissions. This model was also built from a chimera. This model used try2-opt2 of the M1-T64 subdomain, which was a polish of the initial undertaker run for this domain. A65-L224 used try4-opt2 which was based on 2f6sA. L225-T299 used try5-opt2 of this subdomain which attempted to used secondary structure predictions and hydrogen bond predictions for this domain. This model was a polished model of the chimera. ----------