Systematic exploration of ubiquitin sequence, E1 activation efficiency, and experimental fitness in yeast

Zero-shot prediction of mutation effects with multimodal deep representation learning guides protein engineering

Mutations in amino acid sequences can provoke changes in protein function. Accurate and unsupervised prediction of mutation effects is critical in biotechnology and biomedicine, but remains a fundamental challenge. To resolve this challenge, here we present Protein Mutational Effect Predictor (ProMEP), a general and multiple sequence alignment-free method that enables zero-shot prediction of mutation effects. A multimodal deep representation learning model embedded in ProMEP was developed to comprehensively learn both sequence and structure contexts from ~160 million proteins. ProMEP achieves state-of-the-art performance in mutational effect prediction and accomplishes a tremendous improvement in speed, enabling efficient and intelligent protein engineering. Specifically, ProMEP accurately forecasts mutational consequences on the gene-editing enzymes TnpB and TadA, and successfully guides the development of high-performance gene-editing tools with their engineered variants. The gene-editing efficiency of a 5-site mutant of TnpB reaches up to 74.04% (vs 24.66% for the wild type); and the base editing tool developed on the basis of a TadA 15-site mutant (in addition to the A106V/D108N double mutation that renders deoxyadenosine deaminase activity to TadA) exhibits an A-to-G conversion frequency of up to 77.27% (vs 69.80% for ABE8e, a previous TadA-based adenine base editor) with significantly reduced bystander and off-target effects compared to ABE8e. ProMEP not only showcases superior performance in predicting mutational effects on proteins but also demonstrates a great capability to guide protein engineering. Therefore, ProMEP enables efficient exploration of the gigantic protein space and facilitates practical design of proteins, thereby advancing studies in biomedicine and synthetic biology.

Evolutionary Probability and Stacked Regressions Enable Data-Driven Protein Engineering with Minimized Experimental Effort

Protein engineering through directed evolution and (semi)rational approaches is routinely applied to optimize protein properties for a broad range of applications in industry and academia. The multitude of possible variants, combined with limited screening throughput, hampers efficient protein engineering. Data-driven strategies have emerged as a powerful tool to model the protein fitness landscape that can be explored in silico, significantly accelerating protein engineering campaigns. However, such methods require a certain amount of data, which often cannot be provided, to generate a reliable model of the fitness landscape. Here, we introduce MERGE, a method that combines direct coupling analysis (DCA) and machine learning (ML). MERGE enables data-driven protein engineering when only limited data are available for training, typically ranging from 50 to 500 labeled sequences. Our method demonstrates remarkable performance in predicting a protein's fitness value and rank based on its sequence across diverse proteins and properties. Notably, MERGE outperforms state-of-the-art methods when only small data sets are available for modeling, requiring fewer computational resources, and proving particularly promising for protein engineers who have access to limited amounts of data.

Unsupervised evolution of protein and antibody complexes with a structure-informed language model

Large language models trained on sequence information alone can learn high-level principles of protein design. However, beyond sequence, the three-dimensional structures of proteins determine their specific function, activity, and evolvability. Here, we show that a general protein language model augmented with protein structure backbone coordinates can guide evolution for diverse proteins without the need to model individual functional tasks. We also demonstrate that ESM-IF1, which was only trained on single-chain structures, can be extended to engineer protein complexes. Using this approach, we screened about 30 variants of two therapeutic clinical antibodies used to treat severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection. We achieved up to 25-fold improvement in neutralization and 37-fold improvement in affinity against antibody-escaped viral variants of concern BQ.1.1 and XBB.1.5, respectively. These findings highlight the advantage of integrating structural information to identify efficient protein evolution trajectories without requiring any task-specific training data.

Functional effects of mutations in proteins can be predicted and interpreted by guided selection of sequence covariation information

Predicting the effects of one or more mutations to the in vivo or in vitro properties of a wild-type protein is a major computational challenge, due to the presence of epistasis, that is, of interactions between amino acids in the sequence. We introduce a computationally efficient procedure to build minimal epistatic models to predict mutational effects by combining evolutionary (homologous sequence) and few mutational-scan data. Mutagenesis measurements guide the selection of links in a sparse graphical model, while the parameters on the nodes and the edges are inferred from sequence data. We show, on 10 mutational scans, that our pipeline exhibits performances comparable to state-of-the-art deep networks trained on many more data, while requiring much less parameters and being hence more interpretable. In particular, the identified interactions adapt to the wild-type protein and to the fitness or biochemical property experimentally measured, mostly focus on key functional sites, and are not necessarily related to structural contacts. Therefore, our method is able to extract information relevant for one mutational experiment from homologous sequence data reflecting the multitude of structural and functional constraints acting on proteins throughout evolution.

Inverse folding of protein complexes with a structure-informed language model enables unsupervised antibody evolution

Large language models trained on sequence information alone are capable of learning high level principles of protein design. However, beyond sequence, the three-dimensional structures of proteins determine their specific function, activity, and evolvability. Here we show that a general protein language model augmented with protein structure backbone coordinates and trained on the inverse folding problem can guide evolution for diverse proteins without needing to explicitly model individual functional tasks. We demonstrate inverse folding to be an effective unsupervised, structure-based sequence optimization strategy that also generalizes to multimeric complexes by implicitly learning features of binding and amino acid epistasis. Using this approach, we screened ~30 variants of two therapeutic clinical antibodies used to treat SARS-CoV-2 infection and achieved up to 26-fold improvement in neutralization and 37-fold improvement in affinity against antibody-escaped viral variants-of-concern BQ.1.1 and XBB.1.5, respectively. In addition to substantial overall improvements in protein function, we find inverse folding performs with leading experimental success rates among other reported machine learning-guided directed evolution methods, without requiring any task-specific training data.

ProteinGym: Large-Scale Benchmarks for Protein Design and Fitness Prediction

Predicting the effects of mutations in proteins is critical to many applications, from understanding genetic disease to designing novel proteins that can address our most pressing challenges in climate, agriculture and healthcare. Despite a surge in machine learning-based protein models to tackle these questions, an assessment of their respective benefits is challenging due to the use of distinct, often contrived, experimental datasets, and the variable performance of models across different protein families. Addressing these challenges requires scale. To that end we introduce ProteinGym, a large-scale and holistic set of benchmarks specifically designed for protein fitness prediction and design. It encompasses both a broad collection of over 250 standardized deep mutational scanning assays, spanning millions of mutated sequences, as well as curated clinical datasets providing high-quality expert annotations about mutation effects. We devise a robust evaluation framework that combines metrics for both fitness prediction and design, factors in known limitations of the underlying experimental methods, and covers both zero-shot and supervised settings. We report the performance of a diverse set of over 70 high-performing models from various subfields (eg., alignment-based, inverse folding) into a unified benchmark suite. We open source the corresponding codebase, datasets, MSAs, structures, model predictions and develop a user-friendly website that facilitates data access and analysis.

Ensemble Learning with Supervised Methods Based on Large-Scale Protein Language Models for Protein Mutation Effects Prediction

Machine learning has been increasingly utilized in the field of protein engineering, and research directed at predicting the effects of protein mutations has attracted increasing attention. Among them, so far, the best results have been achieved by related methods based on protein language models, which are trained on a large number of unlabeled protein sequences to capture the generally hidden evolutionary rules in protein sequences, and are therefore able to predict their fitness from protein sequences. Although numerous similar models and methods have been successfully employed in practical protein engineering processes, the majority of the studies have been limited to how to construct more complex language models to capture richer protein sequence feature information and utilize this feature information for unsupervised protein fitness prediction. There remains considerable untapped potential in these developed models, such as whether the prediction performance can be further improved by integrating different models to further improve the accuracy of prediction. Furthermore, how to utilize large-scale models for prediction methods of mutational effects on quantifiable properties of proteins due to the nonlinear relationship between protein fitness and the quantification of specific functionalities has yet to be explored thoroughly. In this study, we propose an ensemble learning approach for predicting mutational effects of proteins integrating protein sequence features extracted from multiple large protein language models, as well as evolutionarily coupled features extracted in homologous sequences, while comparing the differences between linear regression and deep learning models in mapping these features to quantifiable functional changes. We tested our approach on a dataset of 17 protein deep mutation scans and indicated that the integrated approach together with linear regression enables the models to have higher prediction accuracy and generalization. Moreover, we further illustrated the reliability of the integrated approach by exploring the differences in the predictive performance of the models across species and protein sequence lengths, as well as by visualizing clustering of ensemble and non-ensemble features.

Accurate prediction of functional effect of single amino acid variants with deep learning

The assessment of functional effect of amino acid variants is a critical biological problem in proteomics for clinical medicine and protein engineering. Although natively occurring variants offer insights into deleterious variants, high-throughput deep mutational experiments enable comprehensive investigation of amino acid variants for a given protein. However, these mutational experiments are too expensive to dissect millions of variants on thousands of proteins. Thus, computational approaches have been proposed, but they heavily rely on hand-crafted evolutionary conservation, limiting their accuracy. Recent advancement in transformers provides a promising solution to precisely estimate the functional effects of protein variants on high-throughput experimental data. Here, we introduce a novel deep learning model, namely Rep2Mut-V2, which leverages learned representation from transformer models. Rep2Mut-V2 significantly enhances the prediction accuracy for 27 types of measurements of functional effects of protein variants. In the evaluation of 38 protein datasets with 118,933 single amino acid variants, Rep2Mut-V2 achieved an average Spearman's correlation coefficient of 0.7. This surpasses the performance of six state-of-the-art methods, including the recently released methods ESM, DeepSequence and EVE. Even with limited training data, Rep2Mut-V2 outperforms ESM and DeepSequence, showing its potential to extend high-throughput experimental analysis for more protein variants to reduce experimental cost. In conclusion, Rep2Mut-V2 provides accurate predictions of the functional effects of single amino acid variants of protein coding sequences. This tool can significantly aid in the interpretation of variants in human disease studies.

Systematic profiling of dominant ubiquitin variants reveals key functional nodes contributing to evolutionary selection

Dominant-negative mutations can help to investigate the biological mechanisms and to understand the selective pressures for multifunctional proteins. However, most studies have focused on recessive mutant effects that occur in the absence of a second functional gene copy, which overlooks the fact that most eukaryotic genomes contain more than one copy of many genes. We have identified dominant effects on yeast growth rate among all possible point mutations in ubiquitin expressed alongside a wild-type allele. Our results reveal more than 400 dominant-negative mutations, indicating that dominant-negative effects make a sizable contribution to selection acting on ubiquitin. Cellular and biochemical analyses of individual ubiquitin variants show that dominant-negative effects are explained by varied accumulation of polyubiquitinated cellular proteins and/or defects in conjugation of ubiquitin variants to ubiquitin ligases. Our approach to identify dominant-negative mutations is general and can be applied to other proteins of interest.

Mutational fitness landscape and drug resistance

Robust technology has been developed to systematically quantify fitness landscapes that provide valuable opportunities to improve our understanding of drug resistance and define new avenues to develop drugs with reduced resistance susceptibility. We outline the critical importance of drug resistance studies and the potential for fitness landscape approaches to contribute to this effort. We describe the major technical advancements in mutational scanning, which is the primary approach used to quantify protein fitness landscapes. There are many complex steps to consider in planning and executing mutational scanning projects including developing a selection scheme, generating mutant libraries, tracking the frequency of variants using next-generation sequencing, and processing and interpreting the data. Key experimental parameters impacting each of these steps are discussed to aid in planning fitness landscape studies. There is a strong need for improved understanding of drug resistance, and fitness landscapes provide a promising new approach.

Integrating deep mutational scanning and low-throughput mutagenesis data to predict the impact of amino acid variants

BACKGROUND

Evaluating the impact of amino acid variants has been a critical challenge for studying protein function and interpreting genomic data. High-throughput experimental methods like deep mutational scanning (DMS) can measure the effect of large numbers of variants in a target protein, but because DMS studies have not been performed on all proteins, researchers also model DMS data computationally to estimate variant impacts by predictors.

RESULTS

In this study, we extended a linear regression-based predictor to explore whether incorporating data from alanine scanning (AS), a widely used low-throughput mutagenesis method, would improve prediction results. To evaluate our model, we collected 146 AS datasets, mapping to 54 DMS datasets across 22 distinct proteins.

CONCLUSIONS

We show that improved model performance depends on the compatibility of the DMS and AS assays, and the scale of improvement is closely related to the correlation between DMS and AS results.

Analyses of HIV proteases variants at the threshold of viability reveals relationships between processing efficiency and fitness

Investigating the relationships between protein function and fitness provides keys for understanding biochemical mechanisms that underly evolution. Mutations with partial fitness defects can delineate the threshold of biochemical function required for viability. We utilized a previous deep mutational scan of HIV-1 protease (PR) to identify variants with 15–45 per cent defects in replication and analysed the biochemical function of eight variants (L10M, L10S, V32C, V32I, A71V, A71S, Q92I, Q92N). We purified each variant and assessed the efficiency of peptide cleavage for three cut sites (MA-CA, TF-PR, and PR-RT) as well as gel-based analyses of processing of purified Gag. The cutting activity of at least one site was perturbed relative to WT protease for all variants, consistent with cutting activity being a primary determinant of fitness effects. We examined the correlation of fitness defects with cutting activity of different sites. MA-CA showed the weakest correlation (R² = 0.02) with fitness, suggesting relatively weak coupling with viral replication. In contrast, cutting of the TF-PR site showed the strongest correlation with fitness (R² = 0.53). Cutting at the TF-PR site creates a new PR protein with a free N-terminus that is critical for activity. Our findings indicate that increasing the pool of active PR is rate limiting for viral replication, making this an ideal step to target with inhibitors.

AMaLa: Analysis of Directed Evolution Experiments via Annealed Mutational Approximated Landscape

We present Annealed Mutational approximated Landscape (AMaLa), a new method to infer fitness landscapes from Directed Evolution experiments sequencing data. Such experiments typically start from a single wild-type sequence, which undergoes Darwinian in vitro evolution via multiple rounds of mutation and selection for a target phenotype. In the last years, Directed Evolution is emerging as a powerful instrument to probe fitness landscapes under controlled experimental conditions and as a relevant testing ground to develop accurate statistical models and inference algorithms (thanks to high-throughput screening and sequencing). Fitness landscape modeling either uses the enrichment of variants abundances as input, thus requiring the observation of the same variants at different rounds or assuming the last sequenced round as being sampled from an equilibrium distribution. AMaLa aims at effectively leveraging the information encoded in the whole time evolution. To do so, while assuming statistical sampling independence between sequenced rounds, the possible trajectories in sequence space are gauged with a time-dependent statistical weight consisting of two contributions: (i) an energy term accounting for the selection process and (ii) a generalized Jukes-Cantor model for the purely mutational step. This simple scheme enables accurately describing the Directed Evolution dynamics and inferring a fitness landscape that correctly reproduces the measures of the phenotype under selection (e.g., antibiotic drug resistance), notably outperforming widely used inference strategies. In addition, we assess the reliability of AMaLa by showing how the inferred statistical model could be used to predict relevant structural properties of the wild-type sequence.

Exploring amino acid functions in a deep mutational landscape

Amino acids fulfil a diverse range of roles in proteins, each utilising its chemical properties in different ways in different contexts to create required functions. For example, cysteines form disulphide or hydrogen bonds in different circumstances and charged amino acids do not always make use of their charge. The repertoire of amino acid functions and the frequency at which they occur in proteins remains understudied. Measuring large numbers of mutational consequences, which can elucidate the role an amino acid plays, was prohibitively time-consuming until recent developments in deep mutational scanning. In this study, we gathered data from 28 deep mutational scanning studies, covering 6,291 positions in 30 proteins, and used the consequences of mutation at each position to define a mutational landscape. We demonstrated rich relationships between this landscape and biophysical or evolutionary properties. Finally, we identified 100 functional amino acid subtypes with a data-driven clustering analysis and studied their features, including their frequencies and chemical properties such as tolerating polarity, hydrophobicity or being intolerant of charge or specific amino acids. The mutational landscape and amino acid subtypes provide a foundational catalogue of amino acid functional diversity, which will be refined as the number of studied protein positions increases.

Evotuning protocols for Transformer-based variant effect prediction on multi-domain proteins

Accurate variant effect prediction has broad impacts on protein engineering. Recent machine learning approaches toward this end are based on representation learning, by which feature vectors are learned and generated from unlabeled sequences. However, it is unclear how to effectively learn evolutionary properties of an engineering target protein from homologous sequences, taking into account the protein's sequence-level structure called domain architecture (DA). Additionally, no optimal protocols are established for incorporating such properties into Transformer, the neural network well-known to perform the best in natural language processing research. This article proposes DA-aware evolutionary fine-tuning, or 'evotuning', protocols for Transformer-based variant effect prediction, considering various combinations of homology search, fine-tuning and sequence vectorization strategies. We exhaustively evaluated our protocols on diverse proteins with different functions and DAs. The results indicated that our protocols achieved significantly better performances than previous DA-unaware ones. The visualizations of attention maps suggested that the structural information was incorporated by evotuning without direct supervision, possibly leading to better prediction accuracy.

Unsupervised Inference of Protein Fitness Landscape from Deep Mutational Scan

The recent technological advances underlying the screening of large combinatorial libraries in high-throughput mutational scans deepen our understanding of adaptive protein evolution and boost its applications in protein design. Nevertheless, the large number of possible genotypes requires suitable computational methods for data analysis, the prediction of mutational effects, and the generation of optimized sequences. We describe a computational method that, trained on sequencing samples from multiple rounds of a screening experiment, provides a model of the genotype-fitness relationship. We tested the method on five large-scale mutational scans, yielding accurate predictions of the mutational effects on fitness. The inferred fitness landscape is robust to experimental and sampling noise and exhibits high generalization power in terms of broader sequence space exploration and higher fitness variant predictions. We investigate the role of epistasis and show that the inferred model provides structural information about the 3D contacts in the molecular fold.

The Roles of Protein Structure, Taxon Sampling, and Model Complexity in Phylogenomics: A Case Study Focused on Early Animal Divergences

Despite the long history of using protein sequences to infer the tree of life, the potential for different parts of protein structures to retain historical signal remains unclear. We propose that it might be possible to improve analyses of phylogenomic datasets by incorporating information about protein structure. We test this idea using the position of the root of Metazoa (animals) as a model system. We examined the distribution of “strongly decisive” sites (alignment positions that support a specific tree topology) in a dataset comprising >1500 proteins and almost 100 taxa. The proportion of each class of strongly decisive sites in different structural environments was very sensitive to the model used to analyze the data when a limited number of taxa were used but they were stable when taxa were added. As long as enough taxa were analyzed, sites in all structural environments supported the same topology regardless of whether standard tree searches or decisive sites were used to select the optimal tree. However, the use of decisive sites revealed a difference between the support for minority topologies for sites in different structural environments: buried sites and sites in sheet and coil environments exhibited equal support for the minority topologies, whereas solvent-exposed and helix sites had unequal numbers of sites, supporting the minority topologies. This suggests that the relatively slowly evolving buried, sheet, and coil sites are giving an accurate picture of the true species tree and the amount of conflict among gene trees. Taken as a whole, this study indicates that phylogenetic analyses using sites in different structural environments can yield different topologies for the deepest branches in the animal tree of life and that analyzing larger numbers of taxa eliminates this conflict. More broadly, our results highlight the desirability of incorporating information about protein structure into phylogenomic analyses.

A large-scale survey of pairwise epistasis reveals a mechanism for evolutionary expansion and specialization of PDZ domains

Deep mutational scanning (DMS) facilitates data-driven models of protein structure and function. Here, we adapted Saturated Programmable Insertion Engineering (SPINE) as a programmable DMS technique. We validate SPINE with a reference single mutant dataset in the PSD95 PDZ3 domain and then characterize most pairwise double mutants to study epistasis. We observe wide-spread proximal negative epistasis, which we attribute to mutations affecting thermodynamic stability, and strong long-range positive epistasis, which is enriched in an evolutionarily conserved and function-defining network of "sector" and clade-specifying residues. Conditional neutrality of mutations in clade-specifying residues compensates for deleterious mutations in sector positions. This suggests that epistatic interactions between these position pairs facilitated the evolutionary expansion and specialization of PDZ domains. We propose that SPINE provides easy experimental access to reveal epistasis signatures in proteins that will improve our understanding of the structural basis for protein function and adaptation.

Balance between promiscuity and specificity in phage λ host range

As hosts acquire resistance to viruses, viruses must overcome that resistance to re-establish infectivity, or go extinct. Despite the significant hurdles associated with adapting to a resistant host, viruses are evolutionarily successful and maintain stable coevolutionary relationships with their hosts. To investigate the factors underlying how pathogens adapt to their hosts, we performed a deep mutational scan of the region of the λ tail fiber tip protein that mediates contact with the receptor on λ's host, Escherichia coli. Phages harboring amino acid substitutions were subjected to selection for infectivity on wild type E. coli, revealing a highly restrictive fitness landscape, in which most substitutions completely abrogate function. A subset of positions that are tolerant of mutation in this assay, but diverse over evolutionary time, are associated with host range expansion. Imposing selection for phage infectivity on three λ-resistant hosts, each harboring a different missense mutation in the λ receptor, reveals hundreds of adaptive variants in λ. We distinguish λ variants that confer promiscuity, a general ability to overcome host resistance, from those that drive host-specific infectivity. Both processes may be important in driving adaptation to a novel host.

ELASPIC2 (EL2): Combining Contextualized Language Models and Graph Neural Networks to Predict Effects of Mutations

The ELASPIC web server allows users to evaluate the effect of mutations on protein folding and protein-protein interaction on a proteome-wide scale. It uses homology models of proteins and protein-protein interactions, which have been precalculated for several proteomes, and machine learning models, which integrate structural information with sequence conservation scores, in order to make its predictions. Since the original publication of the ELASPIC web server, several advances have motivated a revisiting of the problem of mutation effect prediction. First, progress in neural network architectures and self-supervised pre-trained has resulted in models which provide more informative embeddings of protein sequence and structure than those used by the original version of ELASPIC. Second, the amount of training data has increased several-fold, largely driven by advances in deep mutation scanning and other multiplexed assays of variant effect. Here, we describe two machine learning models which leverage the recent advances in order to achieve superior accuracy in predicting the effect of mutation on protein folding and protein-protein interaction. The models incorporate features generated using pre-trained transformer- and graph convolution-based neural networks, and are trained to optimize a ranking objective function, which permits the use of heterogeneous training data. The outputs from the new models have been incorporated into the ELASPIC web server, available at http://elaspic.kimlab.org.

DeMaSk: a deep mutational scanning substitution matrix and its use for variant impact prediction

Motivation

Accurately predicting the quantitative impact of a substitution on a protein’s molecular function would be a great aid in understanding the effects of observed genetic variants across populations. While this remains a challenging task, new approaches can leverage data from the increasing numbers of comprehensive deep mutational scanning (DMS) studies that systematically mutate proteins and measure fitness.

Results

We introduce DeMaSk, an intuitive and interpretable method based only upon DMS datasets and sequence homologs that predicts the impact of missense mutations within any protein. DeMaSk first infers a directional amino acid substitution matrix from DMS datasets and then fits a linear model that combines these substitution scores with measures of per-position evolutionary conservation and variant frequency across homologs. Despite its simplicity, DeMaSk has state-of-the-art performance in predicting the impact of amino acid substitutions, and can easily and rapidly be applied to any protein sequence.

Availability and implementation

https://demask.princeton.edu generates fitness impact predictions and visualizations for any user-submitted protein sequence.

Supplementary information

Supplementary data are available at Bioinformatics online.

Phylogenetic Analyses of Sites in Different Protein Structural Environments Result in Distinct Placements of the Metazoan Root

Phylogenomics, the use of large datasets to examine phylogeny, has revolutionized the study of evolutionary relationships. However, genome-scale data have not been able to resolve all relationships in the tree of life; this could reflect, at least in part, the poor-fit of the models used to analyze heterogeneous datasets. Some of the heterogeneity may reflect the different patterns of selection on proteins based on their structures. To test that hypothesis, we developed a pipeline to divide phylogenomic protein datasets into subsets based on secondary structure and relative solvent accessibility. We then tested whether amino acids in different structural environments had distinct signals for the topology of the deepest branches in the metazoan tree. We focused on a dataset that appeared to have a mixture of signals and we found that the most striking difference in phylogenetic signal reflected relative solvent accessibility. Analyses of exposed sites (residues located on the surface of proteins) yielded a tree that placed ctenophores sister to all other animals whereas sites buried inside proteins yielded a tree with a sponge+ctenophore clade. These differences in phylogenetic signal were not ameliorated when we conducted analyses using a set of maximum-likelihood profile mixture models. These models are very similar to the Bayesian CAT model, which has been used in many analyses of deep metazoan phylogeny. In contrast, analyses conducted after recoding amino acids to limit the impact of deviations from compositional stationarity increased the congruence in the estimates of phylogeny for exposed and buried sites; after recoding amino acid trees estimated using the exposed and buried site both supported placement of ctenophores sister to all other animals. Although the central conclusion of our analyses is that sites in different structural environments yield distinct trees when analyzed using models of protein evolution, our amino acid recoding analyses also have implications for metazoan evolution. Specifically, our results add to the evidence that ctenophores are the sister group of all other animals and they further suggest that the placozoa+cnidaria clade found in some other studies deserves more attention. Taken as a whole, these results provide striking evidence that it is necessary to achieve a better understanding of the constraints due to protein structure to improve phylogenetic estimation.

Variant effect predictions capture some aspects of deep mutational scanning experiments

BACKGROUND

Deep mutational scanning (DMS) studies exploit the mutational landscape of sequence variation by systematically and comprehensively assaying the effect of single amino acid variants (SAVs; also referred to as missense mutations, or non-synonymous Single Nucleotide Variants - missense SNVs or nsSNVs) for particular proteins. We assembled SAV annotations from 22 different DMS experiments and normalized the effect scores to evaluate variant effect prediction methods. Three trained on traditional variant effect data (PolyPhen-2, SIFT, SNAP2), a regression method optimized on DMS data (Envision), and a naïve prediction using conservation information from homologs.

RESULTS

On a set of 32,981 SAVs, all methods captured some aspects of the experimental effect scores, albeit not the same. Traditional methods such as SNAP2 correlated slightly more with measurements and better classified binary states (effect or neutral). Envision appeared to better estimate the precise degree of effect. Most surprising was that the simple naïve conservation approach using PSI-BLAST in many cases outperformed other methods. All methods captured beneficial effects (gain-of-function) significantly worse than deleterious (loss-of-function). For the few proteins with multiple independent experimental measurements, experiments differed substantially, but agreed more with each other than with predictions.

CONCLUSIONS

DMS provides a new powerful experimental means of understanding the dynamics of the protein sequence space. As always, promising new beginnings have to overcome challenges. While our results demonstrated that DMS will be crucial to improve variant effect prediction methods, data diversity hindered simplification and generalization.

MaveDB: an open-source platform to distribute and interpret data from multiplexed assays of variant effect

Multiplex assays of variant effect (MAVEs), such as deep mutational scans and massively parallel reporter assays, test thousands of sequence variants in a single experiment. Despite the importance of MAVE data for basic and clinical research, there is no standard resource for their discovery and distribution. Here, we present MaveDB ( https://www.mavedb.org ), a public repository for large-scale measurements of sequence variant impact, designed for interoperability with applications to interpret these datasets. We also describe the first such application, MaveVis, which retrieves, visualizes, and contextualizes variant effect maps. Together, the database and applications will empower the community to mine these powerful datasets.

GEMME: a simple and fast global epistatic model predicting mutational effects

The systematic and accurate description of protein mutational landscapes is a question of utmost importance in biology, bioengineering, and medicine. Recent progress has been achieved by leveraging on the increasing wealth of genomic data and by modeling intersite dependencies within biological sequences. However, state-of-the-art methods remain time consuming. Here, we present Global Epistatic Model for predicting Mutational Effects (GEMME) (www.lcqb.upmc.fr/GEMME), an original and fast method that predicts mutational outcomes by explicitly modeling the evolutionary history of natural sequences. This allows accounting for all positions in a sequence when estimating the effect of a given mutation. GEMME uses only a few biologically meaningful and interpretable parameters. Assessed against 50 high- and low-throughput mutational experiments, it overall performs similarly or better than existing methods. It accurately predicts the mutational landscapes of a wide range of protein families, including viral ones and, more generally, of much conserved families. Given an input alignment, it generates the full mutational landscape of a protein in a matter of minutes. It is freely available as a package and a webserver at www.lcqb.upmc.fr/GEMME/.

Inferring protein 3D structure from deep mutation scans

We describe an experimental method of three-dimensional (3D) structure determination that exploits the increasing ease of high-throughput mutational scans. Inspired by the success of using natural, evolutionary sequence covariation to compute protein and RNA folds, we explored whether 'laboratory', synthetic sequence variation might also yield 3D structures. We analyzed five large-scale mutational scans and discovered that the pairs of residues with the largest positive epistasis in the experiments are sufficient to determine the 3D fold. We show that the strongest epistatic pairings from genetic screens of three proteins, a ribozyme and a protein interaction reveal 3D contacts within and between macromolecules. Using these experimental epistatic pairs, we compute ab initio folds for a GB1 domain (within 1.8 Å of the crystal structure) and a WW domain (2.1 Å). We propose strategies that reduce the number of mutants needed for contact prediction, suggesting that genomics-based techniques can efficiently predict 3D structure.

Deep generative models of genetic variation capture the effects of mutations

The functions of proteins and RNAs are defined by the collective interactions of many residues, and yet most statistical models of biological sequences consider sites nearly independently. Recent approaches have demonstrated benefits of including interactions to capture pairwise covariation, but leave higher-order dependencies out of reach. Here we show how it is possible to capture higher-order, context-dependent constraints in biological sequences via latent variable models with nonlinear dependencies. We found that DeepSequence ( https://github.com/debbiemarkslab/DeepSequence ), a probabilistic model for sequence families, predicted the effects of mutations across a variety of deep mutational scanning experiments substantially better than existing methods based on the same evolutionary data. The model, learned in an unsupervised manner solely on the basis of sequence information, is grounded with biologically motivated priors, reveals the latent organization of sequence families, and can be used to explore new parts of sequence space.

Multiplexed assays of variant effects contribute to a growing genotype-phenotype atlas

Given the constantly improving cost and speed of genome sequencing, it is reasonable to expect that personal genomes will soon be known for many millions of humans. This stands in stark contrast with our limited ability to interpret the sequence variants which we find. Although it is, perhaps, easiest to interpret variants in coding regions, knowledge of functional impact is unknown for the vast majority of missense variants. While many computational approaches can predict the impact of coding variants, they are given a little weight in the current guidelines for interpreting clinical variants. Laboratory assays produce comparatively more trustworthy results, but until recently did not scale to the space of all possible mutations. The development of deep mutational scanning and other multiplexed assays of variant effect has now brought feasibility of this endeavour within view. Here, we review progress in this field over the last decade, break down the different approaches into their components, and compare methodological differences.

An invisible ubiquitin conformation is required for efficient phosphorylation by PINK1

The Ser/Thr protein kinase PINK1 phosphorylates the well-folded, globular protein ubiquitin (Ub) at a relatively protected site, Ser65. We previously showed that Ser65 phosphorylation results in a conformational change in which Ub adopts a dynamic equilibrium between the known, common Ub conformation and a distinct, second conformation wherein the last β-strand is retracted to extend the Ser65 loop and shorten the C-terminal tail. We show using chemical exchange saturation transfer (CEST) nuclear magnetic resonance experiments that a similar, C-terminally retracted (Ub-CR) conformation also exists at low population in wild-type Ub. Point mutations in the moving β5 and neighbouring β-strands shift the Ub/Ub-CR equilibrium. This enabled functional studies of the two states, and we show that while the Ub-CR conformation is defective for conjugation, it demonstrates improved binding to PINK1 through its extended Ser65 loop, and is a superior PINK1 substrate. Together our data suggest that PINK1 utilises a lowly populated yet more suitable Ub-CR conformation of Ub for efficient phosphorylation. Our findings could be relevant for many kinases that phosphorylate residues in folded protein domains.

Analysis of Large-Scale Mutagenesis Data To Assess the Impact of Single Amino Acid Substitutions

Mutagenesis is a widely used method for identifying protein positions that are important for function or ligand binding. Advances in high-throughput DNA sequencing and mutagenesis techniques have enabled measurement of the effects of nearly all possible amino acid substitutions in many proteins. The resulting large-scale mutagenesis data sets offer a unique opportunity to draw general conclusions about the effects of different amino acid substitutions. Thus, we analyzed 34,373 mutations in 14 proteins whose effects were measured using large-scale mutagenesis approaches. Methionine was the most tolerated substitution, while proline was the least tolerated. We found that several substitutions, including histidine and asparagine, best recapitulated the effects of other substitutions, even when the identity of the wild-type amino acid was considered. The effects of histidine and asparagine substitutions also correlated best with the effects of other substitutions in different structural contexts. Furthermore, highly disruptive substitutions like aspartic and glutamic acid had the most discriminatory power for detecting ligand interface positions. Our work highlights the utility of large-scale mutagenesis data, and our conclusions can help guide future single substitution mutational scans.

Mutation effects predicted from sequence co-variation

Many high-throughput experimental technologies have been developed to assess the effects of large numbers of mutations (variation) on phenotypes. However, designing functional assays for these methods is challenging, and systematic testing of all combinations is impossible, so robust methods to predict the effects of genetic variation are needed. Most prediction methods exploit evolutionary sequence conservation but do not consider the interdependencies of residues or bases. We present EVmutation, an unsupervised statistical method for predicting the effects of mutations that explicitly captures residue dependencies between positions. We validate EVmutation by comparing its predictions with outcomes of high-throughput mutagenesis experiments and measurements of human disease mutations and show that it outperforms methods that do not account for epistasis. EVmutation can be used to assess the quantitative effects of mutations in genes of any organism. We provide pre-computed predictions for ∼7,000 human proteins at http://evmutation.org/.

A Statistical Guide to the Design of Deep Mutational Scanning Experiments

The characterization of the distribution of mutational effects is a key goal in evolutionary biology. Recently developed deep-sequencing approaches allow for accurate and simultaneous estimation of the fitness effects of hundreds of engineered mutations by monitoring their relative abundance across time points in a single bulk competition. Naturally, the achievable resolution of the estimated fitness effects depends on the specific experimental setup, the organism and type of mutations studied, and the sequencing technology utilized, among other factors. By means of analytical approximations and simulations, we provide guidelines for optimizing time-sampled deep-sequencing bulk competition experiments, focusing on the number of mutants, the sequencing depth, and the number of sampled time points. Our analytical results show that sampling more time points together with extending the duration of the experiment improves the achievable precision disproportionately compared with increasing the sequencing depth or reducing the number of competing mutants. Even if the duration of the experiment is fixed, sampling more time points and clustering these at the beginning and the end of the experiment increase experimental power and allow for efficient and precise assessment of the entire range of selection coefficients. Finally, we provide a formula for calculating the 95%-confidence interval for the measurement error estimate, which we implement as an interactive web tool. This allows for quantification of the maximum expected a priori precision of the experimental setup, as well as for a statistical threshold for determining deviations from neutrality for specific selection coefficient estimates.

Determination of ubiquitin fitness landscapes under different chemical stresses in a classroom setting

Ubiquitin is essential for eukaryotic life and varies in only 3 amino acid positions between yeast and humans. However, recent deep sequencing studies indicate that ubiquitin is highly tolerant to single mutations. We hypothesized that this tolerance would be reduced by chemically induced physiologic perturbations. To test this hypothesis, a class of first year UCSF graduate students employed deep mutational scanning to determine the fitness landscape of all possible single residue mutations in the presence of five different small molecule perturbations. These perturbations uncover 'shared sensitized positions' localized to areas around the hydrophobic patch and the C-terminus. In addition, we identified perturbation specific effects such as a sensitization of His68 in HU and a tolerance to mutation at Lys63 in DTT. Our data show how chemical stresses can reduce buffering effects in the ubiquitin proteasome system. Finally, this study demonstrates the potential of lab-based interdisciplinary graduate curriculum.

DOI:http://dx.doi.org/10.7554/eLife.15802.001

The ability of an organism to grow and reproduce, that is, it’s “fitness”, is determined by how its genes interact with the environment. Yeast is a model organism in which researchers can control the exact mutations present in the yeast’s genes (its genotype) and the conditions in which the yeast cells live (their environment). This allows researchers to measure how a yeast cell’s genotype and environment affect its fitness.

Ubiquitin is a protein that many organisms depend on to manage cell stress by acting as a tag that targets other proteins for degradation. Essential proteins such as ubiquitin often remain unchanged by mutation over long periods of time. As a result, these proteins evolve very slowly. Like all proteins, ubiquitin is built from a chain of amino acid molecules linked together, and the ubiquitin proteins of yeast and humans are made of almost identical sequences of amino acids.

Although ubiquitin has barely changed its sequence over evolution, previous studies have shown that – under normal growth conditions in the laboratory – most amino acids in ubiquitin can be mutated without any loss of cell fitness. This led Mavor et al. to hypothesize that treating the yeast cells with chemicals that cause cell stress might lead to amino acids in ubiquitin becoming more sensitive to mutation.

To test this idea, a class of graduate students at the University of California, San Francisco grew yeast cells with different ubiquitin mutations together, and with different chemicals that induce cell stress, and measured their growth rates. Sequencing the ubiquitin gene in the thousands of tested yeast cells revealed that three of the chemicals cause a shared set of amino acids in ubiquitin to become more sensitive to mutation.

This result suggests that these amino acids are important for the stress response, possibly by altering the ability of yeast cells to target certain proteins for degradation. Conversely, another chemical causes yeast to become more tolerant to changes in the ubiquitin sequence. The experiments also link changes in particular amino acids in ubiquitin to specific stress responses.

Mavor et al. show that many of ubquitin’s amino acids are sensitive to mutation under different stress conditions, while others can be mutated to form different amino acids without effecting fitness. By testing the effects of other chemicals, future experiments could further characterize how the yeast’s genotype and environment interact.

DOI:http://dx.doi.org/10.7554/eLife.15802.002

Improving Viral Protease Inhibitors to Counter Drug Resistance

Drug resistance is a major problem in health care, undermining therapy outcomes and necessitating novel approaches to drug design. Extensive studies on resistance to viral protease inhibitors, particularly those of HIV-1 and hepatitis C virus (HCV) protease, revealed a plethora of information on the structural and molecular mechanisms underlying resistance. These insights led to several strategies to improve viral protease inhibitors to counter resistance, such as exploiting the essential biological function and leveraging evolutionary constraints. Incorporation of these strategies into structure-based drug design can minimize vulnerability to resistance, not only for viral proteases but for other quickly evolving drug targets as well, toward designing inhibitors one step ahead of evolution to counter resistance with more intelligent and rational design.

A Balance between Inhibitor Binding and Substrate Processing Confers Influenza Drug Resistance

The therapeutic benefits of the neuraminidase (NA) inhibitor oseltamivir are dampened by the emergence of drug resistance mutations in influenza A virus (IAV). To investigate the mechanistic features that underlie resistance, we developed an approach to quantify the effects of all possible single-nucleotide substitutions introduced into important regions of NA. We determined the experimental fitness effects of 450 nucleotide mutations encoding positions both surrounding the active site and at more distant sites in an N1 strain of IAV in the presence and absence of oseltamivir. NA mutations previously known to confer oseltamivir resistance in N1 strains, including H275Y and N295S, were adaptive in the presence of drug, indicating that our experimental system captured salient features of real-world selection pressures acting on NA. We identified mutations, including several at position 223, that reduce the apparent affinity for oseltamivir in vitro. Position 223 of NA is located adjacent to a hydrophobic portion of oseltamivir that is chemically distinct from the substrate, making it a hotspot for substitutions that preferentially impact drug binding relative to substrate processing. Furthermore, two NA mutations, K221N and Y276F, each reduce susceptibility to oseltamivir by increasing NA activity without altering drug binding. These results indicate that competitive expansion of IAV in the face of drug pressure is mediated by a balance between inhibitor binding and substrate processing.

Viewing protein fitness landscapes through a next-gen lens

High-throughput sequencing has enabled many powerful approaches in biological research. Here, we review sequencing approaches to measure frequency changes within engineered mutational libraries subject to selection. These analyses can provide direct estimates of biochemical and fitness effects for all individual mutations across entire genes (and likely compact genomes in the near future) in genetically tractable systems such as microbes, viruses, and mammalian cells. The effects of mutations on experimental fitness can be assessed using sequencing to monitor time-dependent changes in mutant frequency during bulk competitions. The impact of mutations on biochemical functions can be determined using reporters or other means of separating variants based on individual activities (e.g., binding affinity for a partner molecule can be interrogated using surface display of libraries of mutant proteins and isolation of bound and unbound populations). The comprehensive investigation of mutant effects on both biochemical function and experimental fitness provide promising new avenues to investigate the connections between biochemistry, cell physiology, and evolution. We summarize recent findings from systematic mutational analyses; describe how they relate to a field rich in both theory and experimentation; and highlight how they may contribute to ongoing and future research into protein structure-function relationships, systems-level descriptions of cell physiology, and population-genetic inferences on the relative contributions of selection and drift.

Nonspecific yet decisive: Ubiquitination can affect the native-state dynamics of the modified protein

Ubiquitination is one of the most common post-translational modifications of proteins, and mediates regulated protein degradation among other cellular processes. A fundamental question regarding the mechanism of protein ubiquitination is whether and how ubiquitin affects the biophysical nature of the modified protein. For some systems, it was shown that the position of ubiquitin within the attachment site is quite flexible and ubiquitin does not specifically interact with its substrate. Nevertheless, it was revealed that polyubiquitination can decrease the thermal stability of the modified protein in a site-specific manner because of alterations of the thermodynamic properties of the folded and unfolded states. In this study, we used detailed atomistic simulations to focus on the molecular effects of ubiquitination on the native structure of the modified protein. As a model, we used Ubc7, which is an E2 enzyme whose in vivo ubiquitination process is well characterized and known to lead to degradation. We found that, despite the lack of specific direct interactions between the ubiquitin moiety and Ubc7, ubiquitination decreases the conformational flexibility of certain regions of the substrate Ubc7 protein, which reduces its entropy and thus destabilizes it. The strongest destabilizing effect was observed for systems in which Lys48-linked tetra-ubiquitin was attached to sites used for in vivo degradation. These results reveal how changes in the configurational entropy of the folded state may modulate the stability of the protein's native state. Overall, our results imply that ubiquitination can modify the biophysical properties of the attached protein in the folded state and that, in some proteins, different ubiquitination sites will lead to different biophysical outcomes. We propose that this destabilizing effect of polyubiquitin on the substrate is linked to the functions carried out by the modification, and in particular, regulatory control of protein half-life through proteasomal degradation.