Re: [modeller_usage] Modeller objective function in the modeling of missing residues
To: "He, Amy" <>, "" <>
Subject: Re: [modeller_usage] Modeller objective function in the modeling of missing residues
From: Modeller Caretaker <>
Date: Mon, 26 Jul 2021 18:22:14 -0700
On 7/20/21 9:01 AM, He, Amy wrote:
1. About the "Modeller objective function" as an output score for each
model: Could you help me understand what exactly the function is for
the modelling of the missing residues? Is it a score for just the
modelled loop region, or does it include the pair potential between
the loop and the existent structure?
The objective function is simply the sum of all restraints, including
the statistical potential, excluded volume, and stereochemistry. It
includes the interaction between the loop and the rest of the structure
in the final phase of sampling (otherwise the loop would pass through
the structure). See
https://salilab.org/modeller/10.1/manual/node499.html and the papers
cited there for more information.
2. About the “md_level” of the loop refinement in
“a.loop.md_level = refine.<option>”: Comparing the “fast” and
the “large_slow” options, the “fast” option gave me models with
better (lower) values of the “Modeller objective function”. Is this
normal for the modeling of missing residues? Should I rather choose
the “fast” option not the “large_slow”?
The md_level controls the amount of annealing done in the protocol.
Generally the defaults are fine here - to get good sampling of
conformational space, it is better to build more models with a fast
md_level, not fewer models with a slow md_level.