Transformer protein language models are unsupervised structure learners
Rao, R. M.; Meier, J.; Sercu, T.; Ovchinnikov, S.; Rives, A.
Show abstract
AO_SCPLOWBSTRACTC_SCPLOWUnsupervised 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.1
Matching journals
The top 7 journals account for 50% of the predicted probability mass.