Protein Structural Alignments From Sequence

Abstract

Computing sequence similarity is a fundamental task in biology, with alignment forming the basis for the annotation of genes and genomes and providing the core data structures for evolutionary analysis. Standard approaches are a mainstay of modern molecular biology and rely on variations of edit distance to obtain explicit alignments between pairs of biological sequences. However, sequence alignment algorithms struggle with remote homology tasks and cannot identify similarities between many pairs of proteins with similar structures and likely homology. Recent work suggests that using machine learning language models can improve remote homology detection. To this end, we introduce DeepBLAST, that obtains explicit alignments from residue embeddings learned from a protein language model integrated into an end-to-end differentiable alignment framework. This approach can be accelerated on the GPU architectures and outperforms conventional sequence alignment techniques in terms of both speed and accuracy when identifying structurally similar proteins.

Learned Embeddings from Deep Learning to Visualize and Predict Protein Sets

“…For researchers contributing new protein LMs, bio_embeddings can provide a unified interface to distribute their work to the community, requiring minimal changes for pipeline consumers to make use of new protein LMs. For researchers contributing downstream uses of protein LMs [e.g., for the visualization of attention maps (Vig et al., 2020), which are most closely related to protein contact maps, or for the alignment of protein sequences (Morton et al., 2020)], the bio_embeddings pipeline provides a flexible approach to incorporate their work and directly extends it to all the LMs supported by bio_embeddings . In the future, as we expect more protein LMs to be developed, the bio_embeddings pipeline could be combined with the TAPE (Rao et al., 2019) evaluation system to provide an intuition for protein LM researchers about the best use of their new representations.…”

Vision, challenges and opportunities for a Plant Cell Atlas

“…On the other hand, difficulty identifying distant homologs that share low sequence similarity could obscure functional conservation. Advancement in sequence alignment methods, such as incorporating structural information, could help to discover these distant homologs ( Morton et al, 2020 ). This will probably be an iterative process, where increased resolution of gene expression at the highly resolved cellular or subcellular level could allow inferences about functional orthology across species in large gene families.…”

End-to-end learning of multiple sequence alignments with differentiable Smith-Waterman

“…The latter is relevant when using the method in a neural network pipeline requiring backprogation. [31] our JAX implementations of smooth Smith-Waterman (green), smooth Needleman-Wunsch (orange) and a naive non-vectorized Needleman-Wunsch (blue). Top plots report time for a forward pass, and the bottom plots report time for a forward and backward pass.…”

“…Later, a differentiable kernel-based method for alignment was introduced [38]. More recently, Morton et al implemented a differentiable version of the Needleman-Wunsch algorithm for global pairwise alignment [34, 31]. Our implementation has several advantages: (i) vectorization makes our code faster (Appendices B.1 and B.4)), (ii) we implemented local alignment and an affine gap penalty (Appendix B.5), and (iii) due to the way gaps are parameterized, the output of [31] can not be interpreted as an expected alignment (Appendix B.3).…”

“…Bepler et al first pretrain a bidirectional RNN language model, then freeze this model and train a downstream model using a pseudo-alignment loss [5]. Similarly, Morton et al use a pretrained language model to parametrize the the alignment scoring function [31]. Their loss, however, is purely supervised based on ground-truth structural alignments.…”