Probabilistic Model to Treat Flexibility in Molecular Contacts

TL;DR

This paper introduces a probabilistic model that incorporates molecular flexibility into contact evaluation, combining statistical and physics-based approaches to improve accuracy in binding strength estimation.

Contribution

It presents a novel method integrating nonlocal contact probabilities with local torque-based interactions, enhancing transferability and computational efficiency.

Findings

Probabilistic contact overlap quantifies interaction strength.

Model effectively accounts for structural flexibility.

Feasibility demonstrated on lysine dipeptide structure.

Abstract

Evaluating accessible conformational space is computationally expensive and thermal motions are partly neglected in computer models of molecular interactions. This produces error into the estimates of binding strength. We introduce a method for modelling interactions so that structural flexibility is inherently taken into account. It has a statistical model for 3D properties of 'nonlocal' contacts and a physics based description of 'local' interactions, based on mechanical torque. The form of the torque barrier is derived using a representation of the local electronic structure, which is presumed to improve transferability, compared to traditional force fields. The nonlocal contacts are more distant than 1-4 interactions and Target-atoms are represented by 3D probability densities. Probability mass quantifies strength of contact and is calculated as an overlap integral. Repulsion is described by negative probability density, allowing probability mass to be used as the descriptor of contact preference. As a result, we are able to transform the high-dimensional problem into a simpler evaluation of three-dimensional integrals. We outline how this scoring function gives a tool to study the enthalpy--entropy compensation and demonstrate the feasibility of our approach by evaluating numerical probability masses for chosen side chain to main chain contacts in a lysine dipeptide structure.