Hi, I agree, that dissimilarities in residues will propagate error in the model. But, at the same time more similarities supposed to increase the quality of the model. Similarities and identities between A and B and also between B and C are more than that between A and C. Dissimilarities are as follows: A and B 37% B and C 40% where as A and C 46%
I am not clear how the negative contribution due to dissimilarities (which is less following the process 2) is more than the positive contribution due to similarities(which is more following the process 2) so as to make the model less qualitative than that would be obtained following the process 1. The alignment of C with A obtained from multiple alignment among A, B and C is almost identical to the pairwise alignment between A and C, except one deletion that has been shifted 2 space forward in multiple alignment. If that's the case why the model of C obtained following the process 2 is worse than that would be obtained following the process 1? Sorry, If I misunderstood anything. Raja ________________________________ From: Nathaniel Echols <[email protected]> To: [email protected] Sent: Tue, 5 January, 2010 9:38:47 AM Subject: Re: [ccp4bb] Judging a homology model On Tue, Jan 5, 2010 at 9:24 AM, Raja Dey <[email protected]> wrote: > > > > > > > > > > > > > > > >Assuming all the above data are right, which of the >following way would give the best model for C? >1. >Building homology model of C with A as template. >2. >Building homology model of B with A as template and then use B as >template to build C. >Suppose I used MODELLER for this work where an energy >minimization step is involved after threading the query sequence on the >template. >I need to know which model for C will be better and why? (1). I don't have time to list all of the reasons, but it is a certainty that the dissimilar residues between A and B are not all in the same positions as the dissimilar residues in B and C, and each dissimilar residue is going to result in a less accurate model - so if you model A->B->C, you will be propagating errors. (There is probably a formal mathematical way to express this - there have been many studies on sequence similarity vs. RMSD.) If you need to convince yourself further, find three structures with similar sequence identities in the PDB, run the experiment with them, and compare the end result to the actual structures. From: Raja Dey <[email protected]> ...View Contact To: [email protected] ________________________________ Dear Friends, I have a question about judging a homology model. I have three homologous proteins A, B and C of which only A has 3D crystal structure available. Their similarities/identities are given below. Pair-wise alignment similarity/identity (%) A and B 63/49 A and C 54/39 B and C 60/46 Assuming all the above data are right, which of the following way would give the best model for C? 1. Building homology model of C with A as template. 2. Building homology model of B with A as template and then use B as template to build C. Suppose I used MODELLER for this work where an energy minimization step is involved after threading the query sequence on the template. I need to know which model for C will be better and why? Thanks for your answer… Happy New Year to you all, Raja The INTERNET now has a personality. YOURS! See your Yahoo! Homepage. http://in.yahoo.com/
