Large language models have recently been shown to develop emergent capabilities with scale, going beyond simple pattern matching to perform higher level reasoning and generate lifelike images and text. While language models trained on protein sequences have been studied at a smaller scale, little is known about what they learn about biology as they are scaled up. In this work we train models up to 15 billion parameters, the largest language models of proteins to be evaluated to date. We find that as models are scaled they learn information enabling the prediction of the three-dimensional structure of a protein at the resolution of individual atoms. We present ESMFold for high accuracy end-to-end atomic level structure prediction directly from the individual sequence of a protein. ESMFold has similar accuracy to AlphaFold2 and RoseTTAFold for sequences with low perplexity that are well understood by the language model. ESMFold inference is an order of magnitude faster than AlphaFold2, enabling exploration of the structural space of metagenomic proteins in practical timescales.

Protein modeling is an increasingly popular area of machine learning research. Semi-supervised learning has emerged as an important paradigm in protein modeling due to the high cost of acquiring supervised protein labels, but the current literature is fragmented when it comes to datasets and standardized evaluation techniques. To facilitate progress in this field, we introduce the Tasks Assessing Protein Embeddings (TAPE), a set of five biologically relevant semi-supervised learning tasks spread across different domains of protein biology. We curate tasks into specific training, validation, and test splits to ensure that each task tests biologically relevant generalization that transfers to real-life scenarios. We benchmark a range of approaches to semi-supervised protein representation learning, which span recent work as well as canonical sequence learning techniques. We find that self-supervised pretraining is helpful for almost all models on all tasks, more than doubling performance in some cases. Despite this increase, in several cases features learned by self-supervised pretraining still lag behind features extracted by state-of-the-art non-neural techniques. This gap in performance suggests a huge opportunity for innovative architecture design and improved modeling paradigms that better capture the signal in biological sequences. TAPE will help the machine learning community focus effort on scientifically relevant problems. Toward this end, all data and code used to run these experiments are available at https://github.com/songlab-cal/tape. * Equal Contribution

Unsupervised contact prediction is central to uncovering physical, structural, and functional constraints for protein structure determination and design. For decades, the predominant approach has been to infer evolutionary constraints from a set of related sequences. In the past year, protein language models have emerged as a potential alternative, but performance has fallen short of state-of-the-art approaches in bioinformatics. In this paper we demonstrate that Transformer attention maps learn contacts from the unsupervised language modeling objective. We find the highest capacity models that have been trained to date already outperform a state-of-the-art unsupervised contact prediction pipeline, suggesting these pipelines can be replaced with a single forward pass of an end-to-end model.