Thu Jul 13 09:18:00 PDT 2006 T0377 Make started Thu Jul 13 09:19:35 PDT 2006 Running on lopez.cse.ucsc.edu Thu Jul 13 15:36:46 PDT 2006 Kevin Karplus HMMs get strong hit to 1zpwX and not much else. BLAST gets strong hit to 2ivyA (100% id), which was released 26 June 2006. I assume that T0377 will be canceled. Thu Jul 13 15:53:19 PDT 2006 Kevin Karplus Interestingly, although 2ivyA was in the dunbrack-pdbaa set, it did not move to the top of the hit list, because of it not being in the template libraries. This may indicate a fairly serious flaw in the protocol for selecting templates. Date: Thu, 13 Jul 2006 16:03:58 -0700 From: Kevin Karplus Subject: T0377 cancellation I noticed that T0377 was the same as 2ivyA before the CASP team announced it. I'm a little worried about the script for combining HMM results, because 2ivyA did *not* come out as the first hit (it was third), even though it was in the dunbrack-pdbaa file. It isn't in the template library (yet, anyway), so that may have hurt the scoring. It only scored a little better than 1zpwX with the t06 w0.5 HMM, but had a bigger difference with the t2k w0.5 HMM. These scores (essentially the only ones available for 2ivyA) were so strong that it should have had an excellent e-value in the combined "best-scores" file, but it got dropped WAY down to e-value 17. That seems like *much* too large a penalty for not being in the template library, and certainly not what I had intended. I suspect a bug in the part of the script that is supposed to detect whether a template is in the template library or not. Kevin ------------------------------------------------------------ Make started Thu Jul 13 17:34:40 PDT 2006 Running on cheep.cse.ucsc.edu I *think* I fixed the bug in the best_scores script (and a couple of other oversights, while I was at it). Let's see if it does any better now. OK, 2ivyA is now second to 1zpwX and only a little behind. Since there is a small penalty for not being in the template library (to favor better templates over antiques), this seems quite reasonable.