Protein language models (pLMs) are powerful predictors of protein structure and function, learning through unsupervised training on millions of protein sequences. pLMs are thought to capture common motifs in protein sequences, but the specifics of pLM features are not well understood. Identifying these features would not only shed light on how pLMs work, but potentially uncover novel protein biologystudying the model to study the biology. Motivated by this, we train sparse autoencoders (SAEs) on the residual stream of a pLM, ESM-2. By characterizing SAE features, we determine that pLMs use a combination of generic features and family-specific features to represent a protein. In addition, we demonstrate how known sequence determinants of properties such as thermostability and subcellular localization can be identified by linear probing of SAE features. For predictive features without known functional associations, we hypothesize their role in unknown mechanisms and provide visualization tools to aid their interpretation. Our study gives a better understanding of the limitations of pLMs, and demonstrates how SAE features can be used to help generate hypotheses for biological mechanisms. We release our code, model weights, and feature visualizer.

Proteins power a vast array of functional processes in living cells. The capability to create new proteins with designed structures and functions would thus enable the engineering of cellular behavior and development of protein-based therapeutics and materials. Structure-based protein design aims to find structures that are designable (can be realized by a protein sequence), novel (have dissimilar geometry from natural proteins), and diverse (span a wide range of geometries). While advances in protein structure prediction have made it possible to predict structures of novel protein sequences, the combinatorially large space of sequences and structures limits the practicality of search-based methods. Generative models provide a compelling alternative, by implicitly learning the low-dimensional structure of complex data distributions. Here, we leverage recent advances in denoising diffusion probabilistic models and equivariant neural networks to develop Genie, a generative model of protein structures that performs discrete-time diffusion using a cloud of oriented reference frames in 3D space. Through in silico evaluations, we demonstrate that Genie generates protein backbones that are more designable, novel, and diverse than existing models. This indicates that Genie is capturing key aspects of the distribution of protein structure space and facilitates protein design with high success rates. Code for generating new proteins and training new versions of Genie is available at https://github.com/aqlaboratory/genie.

Multiple sequence alignments (MSAs) of proteins encode rich biological information and have been workhorses in bioinformatic methods for tasks like protein design and protein structure prediction for decades. Recent breakthroughs like AlphaFold2 that use transformers to attend directly over large quantities of raw MSAs have reaffirmed their importance. Generation of MSAs is highly computationally intensive, however, and no datasets comparable to those used to train AlphaFold2 have been made available to the research community, hindering progress in machine learning for proteins. To remedy this problem, we introduce OpenProteinSet, an open-source corpus of more than 16 million MSAs, associated structural homologs from the Protein Data Bank, and AlphaFold2 protein structure predictions. We have previously demonstrated the utility of OpenProteinSet by successfully retraining AlphaFold2 on it. We expect OpenProteinSet to be broadly useful as training and validation data for 1) diverse tasks focused on protein structure, function, and design and 2) large-scale multimodal machine learning research.

SUMMARY: Computational prediction of protein structure from sequence is broadly viewed as a foundational problem of biochemistry and one of the most difficult challenges in bioinformatics. Once every two years the Critical Assessment of protein Structure Prediction (CASP) experiments are held to assess the state of the art in the field in a blind fashion, by presenting predictor groups with protein sequences whose structures have been solved but have not yet been made publicly available. The first CASP was organized in 1994, and the latest, CASP13, took place last December, when for the first time the industrial laboratory DeepMind entered the competition. DeepMind's entry, AlphaFold, placed first in the Free Modeling (FM) category, which assesses methods on their ability to predict novel protein folds (the Zhang group placed first in the Template-Based Modeling (TBM) category, which assess methods on predicting proteins whose folds are related to ones already in the Protein Data Bank.) DeepMind's success generated significant public interest. Their approach builds on two ideas developed in the academic community during the preceding decade: (i) the use of co-evolutionary analysis to map residue co-variation in protein sequence to physical contact in protein structure, and (ii) the application of deep neural networks to robustly identify patterns in protein sequence and co-evolutionary couplings and convert them into contact maps. In this Letter, we contextualize the significance of DeepMind's entry within the broader history of CASP, relate AlphaFold's methodological advances to prior work, and speculate on the future of this important problem.

BACKGROUND: Rapid progress in deep learning has spurred its application to bioinformatics problems including protein structure prediction and design. In classic machine learning problems like computer vision, progress has been driven by standardized data sets that facilitate fair assessment of new methods and lower the barrier to entry for non-domain experts. While data sets of protein sequence and structure exist, they lack certain components critical for machine learning, including high-quality multiple sequence alignments and insulated training/validation splits that account for deep but only weakly detectable homology across protein space. RESULTS: We created the ProteinNet series of data sets to provide a standardized mechanism for training and assessing data-driven models of protein sequence-structure relationships. ProteinNet integrates sequence, structure, and evolutionary information in programmatically accessible file formats tailored for machine learning frameworks. Multiple sequence alignments of all structurally characterized proteins were created using substantial high-performance computing resources. Standardized data splits were also generated to emulate the difficulty of past CASP (Critical Assessment of protein Structure Prediction) experiments by resetting protein sequence and structure space to the historical states that preceded six prior CASPs. Utilizing sensitive evolution-based distance metrics to segregate distantly related proteins, we have additionally created validation sets distinct from the official CASP sets that faithfully mimic their difficulty. CONCLUSION: ProteinNet represents a comprehensive and accessible resource for training and assessing machine-learned models of protein structure.

BACKGROUND: Molecular interactions between proteins and DNA molecules underlie many cellular processes, including transcriptional regulation, chromosome replication, and nucleosome positioning. Computational analyses of protein-DNA interactions rely on experimental data characterizing known protein-DNA interactions structurally and biochemically. While many databases exist that contain either structural or biochemical data, few integrate these two data sources in a unified fashion. Such integration is becoming increasingly critical with the rapid growth of structural and biochemical data, and the emergence of algorithms that rely on the synthesis of multiple data types to derive computational models of molecular interactions. DESCRIPTION: We have developed an integrated affinity-structure database in which the experimental and quantitative DNA binding affinities of helix-turn-helix proteins are mapped onto the crystal structures of the corresponding protein-DNA complexes. This database provides access to: (i) protein-DNA structures, (ii) quantitative summaries of protein-DNA binding affinities using position weight matrices, and (iii) raw experimental data of protein-DNA binding instances. Critically, this database establishes a correspondence between experimental structural data and quantitative binding affinity data at the single basepair level. Furthermore, we present a novel alignment algorithm that structurally aligns the protein-DNA complexes in the database and creates a unified residue-level coordinate system for comparing the physico-chemical environments at the interface between complexes. Using this unified coordinate system, we compute the statistics of atomic interactions at the protein-DNA interface of helix-turn-helix proteins. We provide an interactive website for visualization, querying, and analyzing this database, and a downloadable version to facilitate programmatic analysis. CONCLUSIONS: This database will facilitate the analysis of protein-DNA interactions and the development of programmatic computational methods that capitalize on integration of structural and biochemical datasets. The database can be accessed at http://ProteinDNA.hms.harvard.edu.