Loop-Diffusion: an equivariant diffusion model for designing and scoring protein loops

TL;DR

Loop-Diffusion is a novel energy-based diffusion model that leverages a comprehensive dataset of protein loops to accurately predict protein functions and recognize binding-enhancing mutations, advancing computational protein design.

Contribution

It introduces Loop-Diffusion, an equivariant diffusion model that learns an energy function from diverse protein loops for functional prediction tasks.

Findings

Achieves state-of-the-art in recognizing binding-enhancing mutations.

Effectively scores TCR-pMHC interfaces.

Generalizes well across protein functional prediction tasks.

Abstract

Predicting protein functional characteristics from structure remains a central problem in protein science, with broad implications from understanding the mechanisms of disease to designing novel therapeutics. Unfortunately, current machine learning methods are limited by scarce and biased experimental data, and physics-based methods are either too slow to be useful, or too simplified to be accurate. In this work, we present Loop-Diffusion, an energy based diffusion model which leverages a dataset of general protein loops from the entire protein universe to learn an energy function that generalizes to functional prediction tasks. We evaluate Loop-Diffusion's performance on scoring TCR-pMHC interfaces and demonstrate state-of-the-art results in recognizing binding-enhancing mutations.