Protein sequence-to-structure learning: Is this the end(-to-end revolution)?

TL;DR

Recent advances in deep learning, including geometric learning, protein language models, and end-to-end architectures, have significantly improved protein structure prediction, reaching near-experimental accuracy and transforming the field.

Contribution

This paper provides an overview and critical analysis of the latest deep learning approaches in protein structure prediction developed in the last two years.

Findings

Deep learning has achieved near-experimental accuracy in CASP14.

Emerging methods include geometric learning and pre-trained language models.

End-to-end differentiable models are now being used for structure prediction.

Abstract

The potential of deep learning has been recognized in the protein structure prediction community for some time, and became indisputable after CASP13. In CASP14, deep learning has boosted the field to unanticipated levels reaching near-experimental accuracy. This success comes from advances transferred from other machine learning areas, as well as methods specifically designed to deal with protein sequences and structures, and their abstractions. Novel emerging approaches include (i) geometric learning, i.e. learning on representations such as graphs, 3D Voronoi tessellations, and point clouds; (ii) pre-trained protein language models leveraging attention; (iii) equivariant architectures preserving the symmetry of 3D space; (iv) use of large meta-genome databases; (v) combinations of protein representations; (vi) and finally truly end-to-end architectures, i.e. differentiable models starting from a sequence and returning a 3D structure. Here, we provide an overview and our opinion of the novel deep learning approaches developed in the last two years and widely used in CASP14.