Variational auto-encoding of protein sequences

By Sam Sinai, Eric D. Kelsic, George M. Church, Martin A. Nowak. 2017

74 citations recorded by the source corpus.

Proteins are responsible for the most diverse set of functions in biology. The ability to extract information from protein sequences and to predict the effects of mutations is extremely valuable in many domains of biology and medicine. However the mapping between protein sequence and function is complex and poorly understood. Here we present an embedding of natural protein sequences using a Variational Auto-Encoder and use it to predict how mutations affect protein function. We use this unsupervised approach to cluster natural variants and learn interactions between sets of positions within a protein. This approach generally performs better than baseline methods that consider no interactions within sequences, and in some cases better than the state-of-the-art approaches that use the inverse-Potts model. This generative model can be used to computationally guide exploration of protein sequence space and to better inform rational and automatic protein design.

Subfields: Molecular Biology

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Molecular Biology

Citations in the corpus

Deciphering protein evolution and fitness landscapes with latent space models (2019, methodology): “39 seemed to imply that the latent space model is not better than DCA in predicting effects of mutations when the number of sequences is small and is slightly better when the number of sequences is large.”
Generative Models for Graph-Based Protein Design (2019, background): “In parallel to the development of structure-based models, there has been considerable work on deep generative models for protein sequences in individual protein families with directed (Riesselman et al., 2018; Sinai et al., 2017) and undirected (Tubiana et al.”
Deep generative models for T cell receptor protein sequences (2019, background): “Recent work on deep generative models of proteins inspired our approach ( Sinai et al., 2017 ; Riesselman et al., 2018 ).”
Machine learning in protein engineering (2018, methodology): “Instead of using neural network models to directly learn the mapping from protein sequence to function, Sinai et al. and Riesselman et al. trained variational autoencoders to learn the distribution of allowed mutations within functional protein families (Sinai et al., 2017; Riesselman et al., 2018).”
AdaLead: A simple and robust adaptive greedy search algorithm for sequence design (2020, methodology): “…better local approximate models φ (cid:48) on already published data (Otwinowski et al., 2018; Alipanahi et al., 2015; Riesselman et al., 2017; Sinai et al., 2017), the more recent work is being done on optimization in this setting (Biswas et al., 2018; Angermueller et al., 2020; Brookes &…”
Learning protein constitutive motifs from sequence data (2018, methodology): “…points: our RBM is an extension of direct-based coupling approaches, requires much less hidden units (15 and 50 times less than, respectively, [51] and [52]), has a simple architecture with two layers carrying sequences and representations, infers interpretable weights with biological…”
Machine learning-assisted directed protein evolution with combinatorial libraries (2019, background): “Nevertheless, continuing to collect full datasets with higher throughput methods such as deep mutational scanning (35) serve as valuable test beds for validating the latest machine-learning algorithms for both regression (36, 37) and design (38) that require more data.”
Leveraging implicit knowledge in neural networks for functional dissection and engineering of proteins (2019, unknown): “6, Supplementary Table 3).”