EvoRator2: Predicting Site-specific Amino Acid Substitutions Based on Protein Structural Information Using Deep Learning.

Multiple sequence alignments (MSAs) are the workhorse of molecular evolution and structural biology research. From MSAs, the amino acids that are tolerated at each site during protein evolution can be inferred. However, little is known regarding the repertoire of tolerated amino acids in proteins when only a few or no sequence homologs are available, such as orphan and de novo designed proteins. Here we present EvoRator2, a deep-learning algorithm trained on over 15,000 protein structures that can predict which amino acids are tolerated at any given site, based exclusively on protein structural information mined from atomic coordinate files. We show that EvoRator2 obtained satisfying results for the prediction of position-weighted scoring matrices (PSSM). We further show that EvoRator2 obtained near state-of-the-art performance on proteins with high quality structures in predicting the effect of mutations in deep mutation scanning (DMS) experiments and that for certain DMS targets, EvoRator2 outperformed state-of-the-art methods. We also show that by combining EvoRator2's predictions with those obtained by a state-of-the-art deep-learning method that accounts for the information in the MSA, the prediction of the effect of mutation in DMS experiments was improved in terms of both accuracy and stability. EvoRator2 is designed to predict which amino-acid substitutions are tolerated in such proteins without many homologous sequences, including orphan or de novo designed proteins. We implemented our approach in the EvoRator web server (https://evorator.tau.ac.il).

EvoRator: Prediction of Residue-level Evolutionary Rates from Protein Structures Using Machine Learning.

Measuring evolutionary rates at the residue level is indispensable for gaining structural and functional insights into proteins. State-of-the-art tools for estimating rates take as input a large set of homologous proteins, a probabilistic model of evolution and a phylogenetic tree. However, a gap exists when only few or no homologous proteins can be found, e.g., orphan proteins. In addition, such tools do not take the three-dimensional (3D) structure of the protein into account. The association between the 3D structure and site-specific rates can be learned using machine-learning regression tools from a cohort of proteins for which both the structure and a large set of homologs exist. Here we present EvoRator, a user-friendly web server that implements a machine-learning regression algorithm to predict site-specific evolutionary rates from protein structures. We show that EvoRator outperforms predictions obtained using traditional physicochemical features, such as relative solvent accessibility and weighted contact number. We also demonstrate the application of EvoRator in three common scenarios that arise in protein evolution research: (1) orphan proteins for which no (or few) homologs exist; (2) When homologous sequences exist, our algorithm contrasts structure-based estimates of the evolutionary rates and the phylogeny-based estimates. This allows detecting sites that are likely conserved due to functional rather than structural constraints; (3) Algorithms that only rely on homologous sequence often fail to accurately measure the evolutionary rates of positions in gapped sequence alignments, which frequently occurs as a result of a clade-specific insertion. Our algorithm makes use of training data and known 3D structure of such gapped positions to predict their evolutionary rates. EvoRator is freely available for all users at: https://evorator.tau.ac.il/.

The evolutionary dynamics that retain long neutral genomic sequences in face of indel deletion bias: a model and its application to human introns.

Insertions and deletions (indels) of short DNA segments are common evolutionary events. Numerous studies showed that deletions occur more often than insertions in both prokaryotes and eukaryotes. It raises the question why neutral sequences are not eradicated from the genome. We suggest that this is due to a phenomenon we term border-induced selection. Accordingly, a neutral sequence is bordered between conserved regions. Deletions occurring near the borders occasionally protrude to the conserved region and are thereby subject to strong purifying selection. Thus, for short neutral sequences, an insertion bias is expected. Here, we develop a set of increasingly complex models of indel dynamics that incorporate border-induced selection. Furthermore, we show that short conserved sequences within the neutrally evolving sequence help explain: (i) the presence of very long sequences; (ii) the high variance of sequence lengths; and (iii) the possible emergence of multimodality in sequence length distributions. Finally, we fitted our models to the human intron length distribution, as introns are thought to be mostly neutral and bordered by conserved exons. We show that when accounting for the occurrence of short conserved sequences within introns, we reproduce the main features, including the presence of long introns and the multimodality of intron distribution.

Harnessing Machine Learning To Unravel Protein Degradation in Escherichia coli.

Degradation of intracellular proteins in Gram-negative bacteria regulates various cellular processes and serves as a quality control mechanism by eliminating damaged proteins. To understand what causes the proteolytic machinery of the cell to degrade some proteins while sparing others, we employed a quantitative pulsed-SILAC (stable isotope labeling with amino acids in cell culture) method followed by mass spectrometry analysis to determine the half-lives for the proteome of exponentially growing Escherichia coli, under standard conditions. We developed a likelihood-based statistical test to find actively degraded proteins and identified dozens of fast-degrading novel proteins. Finally, we used structural, physicochemical, and protein-protein interaction network descriptors to train a machine learning classifier to discriminate fast-degrading proteins from the rest of the proteome, achieving an area under the receiver operating characteristic curve (AUC) of 0.72.IMPORTANCE Bacteria use protein degradation to control proliferation, dispose of misfolded proteins, and adapt to physiological and environmental shifts, but the factors that dictate which proteins are prone to degradation are mostly unknown. In this study, we have used a combined computational-experimental approach to explore protein degradation in E. coli We discovered that the proteome of E. coli is composed of three protein populations that are distinct in terms of stability and functionality, and we show that fast-degrading proteins can be identified using a combination of various protein properties. Our findings expand the understanding of protein degradation in bacteria and have implications for protein engineering. Moreover, as rapidly degraded proteins may play an important role in pathogenesis, our findings may help to identify new potential antibacterial drug targets.

COVID-19 pandemic-related lockdown: response time is more important than its strictness.

The rapid spread of SARS-CoV-2 and its threat to health systems worldwide have led governments to take acute actions to enforce social distancing. Previous studies used complex epidemiological models to quantify the effect of lockdown policies on infection rates. However, these rely on prior assumptions or on official regulations. Here, we use country-specific reports of daily mobility from people cellular usage to model social distancing. Our data-driven model enabled the extraction of lockdown characteristics which were crossed with observed mortality rates to show that: (i) the time at which social distancing was initiated is highly correlated with the number of deaths, r2  = 0.64, while the lockdown strictness or its duration is not as informative; (ii) a delay of 7.49 days in initiating social distancing would double the number of deaths; and (iii) the immediate response has a prolonged effect on COVID-19 death toll.