Re: [modeller_usage] Evaluate protein-peptide complexes using modeller
To: SIMON KØNIG <>,
Subject: Re: [modeller_usage] Evaluate protein-peptide complexes using modeller
From: Modeller Caretaker <>
Date: Tue, 20 Mar 2018 12:37:18 -0700
On 03/20/2018 05:43 AM, SIMON KØNIG wrote:
It has been pointed out to me that the evaluation method in modeller
(molpdf, GA341 and DOPE) are not applicable for models of
protein-peptide complexes.
There is an important difference here between scoring functions and
assessment methods. molpdf is the scoring function used to guide the
optimization of Modeller models - it is not designed for assessment and
should not be used as such.
Many assessment functions are statistical in nature and so it's probably
not a good idea to use them to assess a structure that wasn't
represented in the training set. GA341 and DOPE were both trained on
single chain globular proteins, so no, it's not a good idea to apply
them to protein-peptide complexes. Check out SOAP at
https://salilab.org/SOAP/ instead - there is a SOAP-Peptide potential
that was specially trained for protein-peptide complexes.