PT  - JOURNAL ARTICLE
AU  - Roshan Rao
AU  - Jason Liu
AU  - Robert Verkuil
AU  - Joshua Meier
AU  - John F. Canny
AU  - Pieter Abbeel
AU  - Tom Sercu
AU  - Alexander Rives
TI  - MSA Transformer
AID  - 10.1101/2021.02.12.430858
DP  - 2021 Jan 01
TA  - bioRxiv
PG  - 2021.02.12.430858
4099  - http://biorxiv.org/content/early/2021/08/27/2021.02.12.430858.short
4100  - http://biorxiv.org/content/early/2021/08/27/2021.02.12.430858.full
AB  - Unsupervised protein language models trained across millions of diverse sequences learn structure and function of proteins. Protein language models studied to date have been trained to perform inference from individual sequences. The longstanding approach in computational biology has been to make inferences from a family of evo lutionarily related sequences by fitting a model to each family independently. In this work we combine the two paradigms. We introduce a protein language model which takes as input a set of sequences in the form of a multiple sequence alignment. The model interleaves row and column attention across the input sequences and is trained with a variant of the masked language modeling objective across many protein families. The performance of the model surpasses current state-of-the-art unsupervised structure learning methods by a wide margin, with far greater parameter efficiency than prior state-of-the-art protein language models.Competing Interest StatementThe authors have declared no competing interest.