Deep learning of protein sequence design of protein–protein interactions

Abstract

Motivation As more data of experimentally determined protein structures are becoming available, data-driven models to describe protein sequence-structure relationships become more feasible. Within this space, the amino acid sequence design of protein-protein interactions is still a rather challenging subproblem with very low success rates – yet, it is central to most biological processes. Results We developed an attention-based deep learning model inspired by algorithms used for image-caption assignments to design peptides or protein fragment sequences. Our trained model can be applied for the redesign of natural protein-interfaces or the designed protein interaction fragments. Here, we validate the potential by recapitulating naturally occurring protein-protein interactions including antibody-antigen complexes. The designed interfaces accurately capture essential native interactions and have comparable native-like binding affinities in silico. Furthermore, our model does not need a precise backbone location, making it an attractive tool for working with de novo design of protein-protein interactions. Availability The source code of the method is available at https://github.com/strauchlab/iNNterfaceDesign Supplementary information Supplementary data is available at Bioinformatics online.

Homologous Pairs of Low and High Temperature Originating Proteins Spanning the Known Prokaryotic Universe

“…[25][26][27][28] While new and improved models are continuing to be developed, they do not invariably produce designs that retain tertiary structure and activity at high temperatures. [29][30][31] Strategies which are based on substructures, energy functions, or patterns learned from protein structures represented in the PDB, are limited considering that the majority of proteins are non-thermophilic: only 5% of proteins from the top 25 most populous source organisms are thermophilic. [32][33][34][35] The temperature-dependent nature of enthalpic and entropic forces in the protein means that stability at ambient temperature does not necessarily translate to high-temperature stability.…”

A learnable transition from low temperature to high temperature proteins with neural machine translation

“…[9][10][11][12][13][14][15] a property of interest, including thermal stability, [19,20] zero-shot predictors of the same, [21,22] structure prediction models, [23,24] and sequence design models. [25,26] Existing supervised strategies help rank proteins among a pool of variants after training on a specific thermal stability target, but require time and resource intensive labeled data from the specific protein of interest to be accurate. [27][28][29] Zero-shot predictors remove the need for labeled data by either learning from evolutionary scale sequence datasets or by conditioning on homologs of the protein of interest and impressively achieve some predictive performance on observable properties.…”

Prospection of Peptide Inhibitors of Thrombin from Diverse Origins Using a Machine Learning Pipeline

“…As a cheminformatics model, ML combines chemistry, computer science, and information technology to aid in drug discovery through tasks like virtual screening, library design, and high-throughput screening analysis [10][11][12]. Machine learning algorithms leverage large chemical datasets for predictive modeling and pattern recognition, including the prediction of the properties and activities of peptides based on their sidechains [13][14][15][16]. This integration has accelerated the discovery and design of novel peptides with desired biological activities, opening new avenues for peptide-based drug development.…”