Fold-switching proteins push the boundaries of conformational ensemble prediction

TL;DR

This paper investigates the challenges of predicting conformational ensembles of fold-switching proteins using deep learning, highlighting limitations and proposing directions for improving generalizability in modeling dynamic protein structures.

Contribution

The study analyzes how current deep learning models struggle with fold-switching proteins and suggests potential use cases and future directions for better ensemble prediction.

Findings

DL models often predict based on training-set structures

Current methods have limited generalizability for fold-switching proteins

Identifies scenarios where DL approaches succeed or fail

Abstract

A protein's function depends critically on its conformational ensemble, a collection of energy weighted structures whose balance depends on temperature and environment. Though recent deep learning (DL) methods have substantially advanced predictions of single protein structures, computationally modeling conformational ensembles remains a challenge. Here, we focus on modeling fold-switching proteins, which remodel their secondary and/or tertiary structures and change their functions in response to cellular stimuli. These underrepresented members of the protein universe serve as test cases for a method's generalizability. They reveal that DL models often predict conformational ensembles by association with training-set structures, limiting generalizability. These observations suggest use cases for when DL methods will likely succeed or fail. Developing computational methods that successfully identify new fold-switching proteins from large pools of candidates may advance modeling conformational ensembles more broadly.