Hi Frank, Yes, until CASP7 (back in 2006), I used to like saying that there are many more ways to make a homology model worse than the starting template than to make it better, and that homology modelling programs were very good at finding them! After seeing that at least some models (e.g. from Rosetta) were actually better in CASP7, I had to stop saying that!
It’s not just anecdotal. Even in CASP7, most models were still worse for MR than the best template someone could have found. Randy > On 4 Dec 2020, at 12:22, Frank von Delft <frank.vonde...@cmd.ox.ac.uk> wrote: > > I guess that also means that AlphaFold has learnt the crystal-structure-ness > that older homology methods never achieved - which is why (anecdotally?) a > "better" homology model tended to give worse MR performance than the "worse" > template? > > (Or something like that, I'm parrotting what I remember people (maybe Randy?) > saying long ago about the problems with homology models in MR.) > > > On 04/12/2020 11:57, Adam Simpkin wrote: >> I thought I might be able to add a little to this conversation as I >> performed some MR runs as part of the CASP14 High Accuracy analysis. There >> were 30 targets with reflection data. Of these, AlphaFold2 models could be >> used to directly solve 24 structures after converting >> RMS error predictions to simulated B-factors to aid the MR >> (10.1002/prot.25800). >> >> Some of the models did contain sufficient local errors to impede MR. >> However, we were able to obtain a further 3 solutions by using AMPLE to >> truncate the models based on the per-residue RMS error predictions provided. >> In fact, a moderate truncation in AMPLE improved the quality of the MR >> solution in ~78% that succeeded by removing the few incorrectly models loops >> (typically at lattice interfaces). >> >> A final thing to note was that the 3 structures that didn’t work still >> provided high quality model predictions (GDT_TS of 69, 84 & 83). These >> targets all contained multiple chains in the ASU and one was fairly low >> resolution (>3 Angstroms). Overall though I think the take home is clear, >> these models are really good and when the method or something similar is >> more publicly available I think it will definitely simplify MR for >> troublesome targets. >> >> Best wishes, >> >> Adam >> >> ######################################################################## >> >> To unsubscribe from the CCP4BB list, click the following link: >> https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 >> >> This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing >> list hosted by www.jiscmail.ac.uk, terms & conditions are available at >> https://www.jiscmail.ac.uk/policyandsecurity/ > > ######################################################################## > > To unsubscribe from the CCP4BB list, click the following link: > https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 > > This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing > list hosted by www.jiscmail.ac.uk, terms & conditions are available at > https://www.jiscmail.ac.uk/policyandsecurity/ ----- Randy J. Read Department of Haematology, University of Cambridge Cambridge Institute for Medical Research Tel: +44 1223 336500 The Keith Peters Building Fax: +44 1223 336827 Hills Road E-mail: rj...@cam.ac.uk Cambridge CB2 0XY, U.K. www-structmed.cimr.cam.ac.uk ######################################################################## To unsubscribe from the CCP4BB list, click the following link: https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=CCP4BB&A=1 This message was issued to members of www.jiscmail.ac.uk/CCP4BB, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
