Data-Efficient Protein 3D Geometric Pretraining via Refinement of Diffused Protein Structure Decoy

TL;DR

This paper introduces a 3D geometric pretraining framework for proteins using a novel refinement task, addressing data efficiency and physical realism, leading to competitive downstream task performance.

Contribution

It proposes a unified 3D geometric pretraining approach with a protein-specific refinement task, overcoming sample diversity and modeling challenges with limited data.

Findings

Achieved comparable downstream task performance with less than 1% of data used by SOTA models.

Introduced a physically realistic, data-efficient pretraining method for protein structures.

Demonstrated the effectiveness of the RefineDiff task in protein representation learning.

Abstract

Learning meaningful protein representation is important for a variety of biological downstream tasks such as structure-based drug design. Having witnessed the success of protein sequence pretraining, pretraining for structural data which is more informative has become a promising research topic. However, there are three major challenges facing protein structure pretraining: insufficient sample diversity, physically unrealistic modeling, and the lack of protein-specific pretext tasks. To try to address these challenges, we present the 3D Geometric Pretraining. In this paper, we propose a unified framework for protein pretraining and a 3D geometric-based, data-efficient, and protein-specific pretext task: RefineDiff (Refine the Diffused Protein Structure Decoy). After pretraining our geometric-aware model with this task on limited data(less than 1% of SOTA models), we obtained informative protein representations that can achieve comparable performance for various downstream tasks.