ProteinBERT: a universal deep-learning model of protein sequence and function

TLDR

Self‑supervised deep language modeling has achieved unprecedented success in natural language tasks and has recently been adapted to biological sequences, yet existing models are tailored for text rather than proteins. ProteinBERT is a deep language model specifically designed for proteins. Its pretraining combines language modeling with Gene Ontology annotation prediction and employs a novel architecture featuring local and global representations that enable efficient, end‑to‑end processing of long protein sequences. ProteinBERT attains near state‑of‑the‑art performance—sometimes surpassing competitors—on diverse protein benchmarks while being smaller and faster, and it provides an efficient framework for rapid training of protein predictors even with limited labeled data.

Abstract

Self-supervised deep language modeling has shown unprecedented success across natural language tasks, and has recently been repurposed to biological sequences. However, existing models and pretraining methods are designed and optimized for text analysis. We introduce ProteinBERT, a deep language model specifically designed for proteins. Our pretraining scheme combines language modeling with a novel task of Gene Ontology (GO) annotation prediction. We introduce novel architectural elements that make the model highly efficient and flexible to long sequences. The architecture of ProteinBERT consists of both local and global representations, allowing end-to-end processing of these types of inputs and outputs. ProteinBERT obtains near state-of-the-art performance, and sometimes exceeds it, on multiple benchmarks covering diverse protein properties (including protein structure, post-translational modifications and biophysical attributes), despite using a far smaller and faster model than competing deep-learning methods. Overall, ProteinBERT provides an efficient framework for rapidly training protein predictors, even with limited labeled data.