Retamal, P. | Search bioRxiv

bioRxiv2021

Deep learning model for the prediction and classification of protein toxins across all domains of life

Toxins are widely produced by different organisms to disrupt the physiology of other organisms, and support their own existence. Their study is useful to understand protein evolution, environmental adaptation and survival competition. In-silico predictions of toxic proteins can support empirical frameworks, and help in the safety measurements needed for various industrial related processes. Some in-silico methods are slow, hard to implement or lack taxa representation in their training datasets. Here we present a deep learning model to classify protein toxins, through the use of Convolutional Neural Networks (ConvTOX). ConvTOX is able to accurately identify toxic proteins across the domains of life, with accuracies over 80% for animal and plant toxins, and over 50% for bacterial toxins. Moreover, ConvTOX is able to generalize the identification of differences among toxin types, such as neurotoxins and myotoxins, and to accurately identify structural similarities between different protein toxins. ConvTOX overcomes limitations from previous models by being able to predict toxin proteins from across all domains of life, and by not being limited to only short toxin peptides. Limitations are still clear in terms of lower accuracies for specific phylogenetic groups (such as bacterial toxins), but still this works presents itself as a one step forward for the universal use, classification and study of toxic proteins.

synthetic biology↗

bioRxiv2020

Deep learning enables the design of functional de novo antimicrobial proteins

Protein sequences are highly dimensional and present one of the main problems for the optimization and study of sequence-structure relations. The intrinsic degeneration of protein sequences is hard to follow, but the continued discovery of new protein structures has shown that there is convergence in terms of the possible folds that proteins can adopt, such that proteins with sequence identities lower than 30% may still fold into similar structures. Given that proteins share a set of conserved structural motifs, machine-learning algorithms can play an essential role in the study of sequence-structure relations. Deep-learning neural networks are becoming an important tool in the development of new techniques, such as protein modeling and design, and they continue to gain power as new algorithms are developed and as increasing amounts of data are released every day. Here, we trained a deep-learning model based on previous recurrent neural networks to design analog protein structures using representations learning based on the evolutionary and structural information of proteins. We test the capabilities of this model by creating de novo variants of an antifungal peptide, with sequence identities of 50% or lower relative to the wild-type (WT) peptide. We show by in silico approximations, such as molecular dynamics, that the new variants and the WT peptide can successfully bind to a chitin surface with comparable relative binding energies. These results are supported by in vitro assays, where the de novo designed peptides showed antifungal activity that equaled or exceeded the WT peptide.

bioengineering↗