Apologies if this question has been previously posted on the list.
I modelled a sequence a while ago based on three templates using Modeller- with pairwise sequence identities between target and templates: 45%, 38% and 34% (and positives or conserved substitutions in that order as well).
When I overlaid the final model on top of the three templates individually, I got a C alpha r.m.s.d. value of 3.0 Angs for the 45% ID template, 0.6 Angs for the 38% one and 1.4 Angs for the 34% ID template.
How does Modeller deal with multiple templates? My initial thoughts would be that the use of multiple templates would generally increase model accuracy. Is there any reference for this? Also how do you account for the highest sequence similarity template having the highest r.m.s.d of the lot?
Does Modeller simply average out the templates from the multiple alignment between the sequence and three templates in this case? Do other modelling programs deal with this the same way?
Any comments would be greatly appreciated!
Dinesh Soares PhD Student University of Edinburgh