Sparse autoencoders uncover biologically interpretable features in protein language model representations

Contents

## Significance

Interpreting representations derived from protein language models (PLMs) is crucial for improving trust in the model, explainability, and human–AI collaboration in downstream applications. We leverage sparse autoencoders and transcoders from natural language processing as a way to extract biologically meaningful, interpretable features from both protein-level and amino acid-level representations. Our Gene Ontology Analysis and automated interpretability protocols uncover many sparse features that are strongly associated with specific functional annotations and protein families. We show that the sparse features are more interpretable than PLM neurons. These insights not only enhance our understanding of biological information PLMs encode and provide a pathway for gaining functional meaning from PLMs, but also enable interpretability for downstream tasks that rely on these representations.

## Abstract

Foundation models in biology—particularly protein language models (PLMs)—have enabled ground-breaking predictions in protein structure, function, and beyond. However, the “black-box” nature of these representations limits transparency and explainability, posing challenges for human–AI collaboration and leaving open questions about their human-interpretable features. Here, we leverage sparse autoencoders (SAEs) and a variant, transcoders, from natural language processing to extract, in a completely unsupervised fashion, interpretable sparse features present in both protein-level and amino acid (AA)-level representations from ESM2, a popular PLM. Unlike other approaches such as training probes for features, the extraction of features by the SAE is performed without any supervision. We find that many sparse features extracted from SAEs trained on protein-level representations are tightly associated with Gene Ontology (GO) terms across all levels of the GO hierarchy. We also use Anthropic’s Claude to automate the interpretation of sparse features for both protein-level and AA-level representations and find that many of these features correspond to specific protein families and functions such as the NAD Kinase, IUNH, and the PTH family, as well as proteins involved in methyltransferase activity and in olfactory and gustatory sensory perception. We show that sparse features are more interpretable than ESM2 neurons across all our trained SAEs and transcoders. These findings demonstrate that SAEs offer a promising unsupervised approach for disentangling biologically relevant information present in PLM representations, thus aiding interpretability. This work opens the door to safety, trust, and explainability of PLMs and their applications, and paves the way to extracting meaningful biological insights across increasingly powerful models in the life sciences.

### Sign up for PNAS alerts.

Get alerts for new articles, or get an alert when an article is cited.

[Manage alerts](https://www.pnas.org/action/showPreferences?menuTab=Aler

... (truncated, 98 KB total)