MOTIVATION: Predicting how mutations impact protein biophysical properties remains a significant challenge in computational biology. In recent years, numerous predictors, primarily deep learning models, have been developed to address this problem; however, issues such as their lack of interpretability and limited accuracy persist. RESULTS: We showed that a simple evolutionary score, based on the log-odd ratio of wild-type and mutated residue frequencies in evolutionary related proteins, when scaled by the residue's relative solvent accessibility, performs on par with or slightly outperforms most of the benchmarked predictors, many of which are considerably more complex. The evaluation is performed on mutations from the ProteinGym deep mutational scanning dataset collection, which measures various properties such as stability, activity or fitness. This raises further questions about what these complex models actually learn and highlights their limitations in addressing prediction of mutational landscape. AVAILABILITY AND IMPLEMENTATION: The RSALOR model is available as a user-friendly Python package that can be installed from the PyPI repository. The code is freely available at https://github.com/3BioCompBio/RSALOR.

Understanding the impact of mutations on protein-protein binding affinity is a key objective for a wide range of biotechnological applications and for shedding light on disease-causing mutations, which are often located at protein-protein interfaces. Over the past decade, many computational methods using physics-based and/or machine learning approaches have been developed to predict how protein binding affinity changes upon mutations. They all claim to achieve astonishing accuracy on both training and test sets, with performances on standard benchmarks such as SKEMPI 2.0 that seem overly optimistic. Here we benchmarked eight well-known and well-used predictors and identified their biases and dataset dependencies, using not only SKEMPI 2.0 as a test set but also deep mutagenesis data on the severe acute respiratory syndrome coronavirus 2 spike protein in complex with the human angiotensin-converting enzyme 2. We showed that, even though most of the tested methods reach a significant degree of robustness and accuracy, they suffer from limited generalizability properties and struggle to predict unseen mutations. Interestingly, the generalizability problems are more severe for pure machine learning approaches, while physics-based methods are less affected by this issue. Moreover, undesirable prediction biases toward specific mutation properties, the most marked being toward destabilizing mutations, are also observed and should be carefully considered by method developers. We conclude from our analyses that there is room for improvement in the prediction models and suggest ways to check, assess and improve their generalizability and robustness.

MOTIVATION: The accurate prediction of how mutations change biophysical properties of proteins or RNA is a major goal in computational biology with tremendous impacts on protein design and genetic variant interpretation. Evolutionary approaches such as coevolution can help solving this issue. RESULTS: We present pycofitness, a standalone Python-based software package for the in silico mutagenesis of protein and RNA sequences. It is based on coevolution and, more specifically, on a popular inverse statistical approach, namely direct coupling analysis by pseudo-likelihood maximization. Its efficient implementation and user-friendly command line interface make it an easy-to-use tool even for researchers with no bioinformatics background. To illustrate its strengths, we present three applications in which pycofitness efficiently predicts the deleteriousness of genetic variants and the effect of mutations on protein fitness and thermodynamic stability. AVAILABILITY AND IMPLEMENTATION: https://github.com/KIT-MBS/pycofitness.

Accurate in silico prediction of conformational B-cell epitopes would lead to major improvements in disease diagnostics, drug design and vaccine development. A variety of computational methods, mainly based on machine learning approaches, have been developed in the last decades to tackle this challenging problem. Here, we rigorously benchmarked nine state-of-the-art conformational B-cell epitope prediction webservers, including generic and antibody-specific methods, on a dataset of over 250 antibody-antigen structures. The results of our assessment and statistical analyses show that all the methods achieve very low performances, and some do not perform better than randomly generated patches of surface residues. In addition, we also found that commonly used consensus strategies that combine the results from multiple webservers are at best only marginally better than random. Finally, we applied all the predictors to the SARS-CoV-2 spike protein as an independent case study, and showed that they perform poorly in general, which largely recapitulates our benchmarking conclusions. We hope that these results will lead to greater caution when using these tools until the biases and issues that limit current methods have been addressed, promote the use of state-of-the-art evaluation methodologies in future publications and suggest new strategies to improve the performance of conformational B-cell epitope prediction methods.

MOTIVATION: The SARS-CoV-2 virus has shown a remarkable ability to evolve and spread across the globe through successive waves of variants since the original Wuhan lineage. Despite all the efforts of the last 2 years, the early and accurate prediction of variant severity is still a challenging issue which needs to be addressed to help, for example, the decision of activating COVID-19 plans long before the peak of new waves. Upstream preparation would indeed make it possible to avoid the overflow of health systems and limit the most severe cases. RESULTS: We recently developed SpikePro, a structure-based computational model capable of quickly and accurately predicting the viral fitness of a variant from its spike protein sequence. It is based on the impact of mutations on the stability of the spike protein as well as on its binding affinity for the angiotensin-converting enzyme 2 (ACE2) and for a set of neutralizing antibodies. It yields a precise indication of the virus transmissibility, infectivity, immune escape and basic reproduction rate. We present here an updated version of the model that is now available on an easy-to-use webserver, and illustrate its power in a retrospective study of fitness evolution and reproduction rate of the main viral lineages. SpikePro is thus expected to be great help to assess the fitness of newly emerging SARS-CoV-2 variants in genomic surveillance and viral evolution programs. AVAILABILITY AND IMPLEMENTATION: SpikePro webserver http://babylone.ulb.ac.be/SpikePro/. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: High-throughput experiments are generating ever increasing amounts of various -omics data, so shedding new light on the link between human disorders, their genetic causes and the related impact on protein behavior and structure. While numerous bioinformatics tools now exist that predict which variants in the human exome cause diseases, few tools predict the reasons why they might do so. Yet, understanding the impact of variants at the molecular level is a prerequisite for the rational development of targeted drugs or personalized therapies. RESULTS: We present the updated MutaFrame webserver, which aims to meet this need. It offers two deleteriousness prediction softwares, DEOGEN2 and SNPMuSiC, and is designed for bioinformaticians and medical researchers who want to gain insights into the origins of monogenic diseases. It contains information at two levels for each human protein: its amino acid sequence and its three-dimensional structure; we used the experimental structures whenever available, and modeled structures otherwise. MutaFrame also includes higher-level information, such as protein essentiality and protein-protein interactions. It has a user-friendly interface for the interpretation of results and a convenient visualization system for protein structures, in which the variant positions introduced by the user and other structural information are shown. In this way, MutaFrame aids our understanding of the pathogenic processes caused by single-site mutations and their molecular and contextual interpretation. AVAILABILITY AND IMPLEMENTATION: Mutaframe webserver at http://mutaframe.com/. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Although structured proteins adopt their lowest free energy conformation in physiological conditions, the individual residues are generally not in their lowest free energy conformation. Residues that are stability weaknesses are often involved in functional regions, whereas stability strengths ensure local structural stability. The detection of strengths and weaknesses provides key information to guide protein engineering experiments aiming to modulate folding and various functional processes. RESULTS: We developed the SWOTein predictor which identifies strong and weak residues in proteins on the basis of three types of statistical energy functions describing local interactions along the chain, hydrophobic forces and tertiary interactions. The large-scale analysis of the different types of strengths and weaknesses demonstrated their complementarity and the enhancement of the information they provide. Moreover, a good average correlation was observed between predicted and experimental strengths and weaknesses obtained from native hydrogen exchange data. SWOTein application to three test cases further showed its suitability to predict and interpret strong and weak residues in the context of folding, conformational changes and protein-protein binding. In summary, SWOTein is both fast and accurate and can be applied at small and large scale to analyze and modulate folding and molecular recognition processes. AVAILABILITY: The SWOTein webserver provides the list of predicted strengths and weaknesses and a protein structure visualization tool that facilitates the interpretation of the predictions. It is freely available for academic use at http://babylone.ulb.ac.be/SWOTein/.

MOTIVATION: The solubility of a protein is often decisive for its proper functioning. Lack of solubility is a major bottleneck in high-throughput structural genomic studies and in high-concentration protein production, and the formation of protein aggregates causes a wide variety of diseases. Since solubility measurements are time-consuming and expensive, there is a strong need for solubility prediction tools. RESULTS: We have recently introduced solubility-dependent distance potentials that are able to unravel the role of residue-residue interactions in promoting or decreasing protein solubility. Here, we extended their construction by defining solubility-dependent potentials based on backbone torsion angles and solvent accessibility, and integrated them, together with other structure- and sequence-based features, into a random forest model trained on a set of Escherichia coli proteins with experimental structures and solubility values. We thus obtained the SOLart protein solubility predictor, whose most informative features turned out to be folding free energy differences computed from our solubility-dependent statistical potentials. SOLart performances are very good, with a Pearson correlation coefficient between experimental and predicted solubility values of almost 0.7 both in cross-validation on the training dataset and in an independent set of Saccharomyces cerevisiae proteins. On test sets of modeled structures, only a limited drop in performance is observed. SOLart can thus be used with both high-resolution and low-resolution structures, and clearly outperforms state-of-art solubility predictors. It is available through a user-friendly webserver, which is easy to use by non-expert scientists. AVAILABILITY AND IMPLEMENTATION: The SOLart webserver is freely available at http://babylone.ulb.ac.be/SOLART/. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Motivation: Bioinformatics tools that predict protein stability changes upon point mutations have made a lot of progress in the last decades and have become accurate and fast enough to make computational mutagenesis experiments feasible, even on a proteome scale. Despite these achievements, they still suffer from important issues that must be solved to allow further improving their performances and utilizing them to deepen our insights into protein folding and stability mechanisms. One of these problems is their bias toward the learning datasets which, being dominated by destabilizing mutations, causes predictions to be better for destabilizing than for stabilizing mutations. Results: We thoroughly analyzed the biases in the prediction of folding free energy changes upon point mutations (ΔΔG0) and proposed some unbiased solutions. We started by constructing a dataset Ssym of experimentally measured ΔΔG0s with an equal number of stabilizing and destabilizing mutations, by collecting mutations for which the structure of both the wild-type and mutant protein is available. On this balanced dataset, we assessed the performances of 15 widely used ΔΔG0 predictors. After the astonishing observation that almost all these methods are strongly biased toward destabilizing mutations, especially those that use black-box machine learning, we proposed an elegant way to solve the bias issue by imposing physical symmetries under inverse mutations on the model structure, which we implemented in PoPMuSiCsym. This new predictor constitutes an efficient trade-off between accuracy and absence of biases. Some final considerations and suggestions for further improvement of the predictors are discussed. Supplementary information: Supplementary data are available at Bioinformatics online. Note: The article 10.1093/bioinformatics/bty340/, published alongside this paper, also addresses the problem of biases in protein stability change predictions.

MOTIVATION: The molecular bases of protein stability remain far from elucidated even though substantial progress has been made through both computational and experimental investigations. One of the most challenging goals is the development of accurate prediction tools of the temperature dependence of the standard folding free energy ΔG(T). Such predictors have an enormous series of potential applications, which range from drug design in the biopharmaceutical sector to the optimization of enzyme activity for biofuel production. There is thus an important demand for novel, reliable and fast predictors. RESULTS: We present the SCooP algorithm, which is a significant step towards accurate temperature-dependent stability prediction. This automated tool uses the protein structure and the host organism as sole entries and predicts the full T-dependent stability curve of monomeric proteins assumed to follow a two-state folding transition. Equivalently, it predicts all the thermodynamic quantities associated to the folding transition, namely the melting temperature Tm, the standard folding enthalpy ΔHm measured at Tm, and the standard folding heat capacity ΔCp. The cross-validated performances are good, with correlation coefficients between predicted and experimental values equal to [0.80, 0.83, 0.72] for ΔHm, ΔCp and Tm, respectively, which increase up to [0.88, 0.90, 0.78] upon the removal of 10% outliers. Moreover, the stability curve prediction of a target protein is very fast: it takes less than a minute. SCooP can thus potentially be applied on a structurome scale. This opens new perspectives of large-scale analyses of protein stability, which is of considerable interest for protein engineering. AVAILABILITY AND IMPLEMENTATION: The SCooP webserver is freely available at http://babylone.ulb.ac.be/SCooP. CONTACT: [email protected], [email protected] or [email protected]. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: The identification of immunogenic regions on the surface of antigens, which are able to be recognized by antibodies and to trigger an immune response, is a major challenge for the design of new and effective vaccines. The prediction of such regions through computational immunology techniques is a challenging goal, which will ultimately lead to a drastic limitation of the experimental tests required to validate their efficiency. However, current methods are far from being sufficiently reliable and/or applicable on a large scale. RESULTS: We developed SEPIa, a B-cell epitope predictor from the protein sequence, which is sufficiently fast to be applicable on a large scale. The originality of SEPIa lies in the combination of two classifiers, a naïve Bayesian and a random forest classifier, through a voting algorithm that exploits the advantages of both. It is based on 13 sequence-based features, whose values in a 9-residue sequence window are compiled to predict the epitope/non-epitope state of the central residue. The features are related to the type of amino acid, its conservation in homologous proteins, and its tendency of being exposed to the solvent, soluble, flexible, and disordered. The highest signal is obtained from statistical amino acid preferences, but all 13 features contribute non-negligibly in the predictor. SEPIa's average prediction accuracy is limited, with an AUC score (area under the receiver operating characteristic curve) that reaches 0.65 both in 10-fold cross-validation and on an independent test set. It is nevertheless slightly higher than that of other methods evaluated on the same test set. CONCLUSIONS: SEPIa was applied to a test protein whose epitopes are known, human β2 adrenergic G-protein-coupled receptor, with promising results. Although the actual AUC score is rather low, many of the predicted epitopes cluster together and overlap the experimental epitope region. The reasons underlying the limitations of SEPIa and of all other B-cell epitope predictors are discussed.

MOTIVATION: There are now many predictors capable of identifying the likely phenotypic effects of single nucleotide variants (SNVs) or short in-frame Insertions or Deletions (INDELs) on the increasing amount of genome sequence data. Most of these predictors focus on SNVs and use a combination of features related to sequence conservation, biophysical, and/or structural properties to link the observed variant to either neutral or disease phenotype. Despite notable successes, the mapping between genetic variants and their phenotypic effects is riddled with levels of complexity that are not yet fully understood and that are often not taken into account in the predictions, despite their promise of significantly improving the prediction of deleterious mutants. RESULTS: We present DEOGEN, a novel variant effect predictor that can handle both missense SNVs and in-frame INDELs. By integrating information from different biological scales and mimicking the complex mixture of effects that lead from the variant to the phenotype, we obtain significant improvements in the variant-effect prediction results. Next to the typical variant-oriented features based on the evolutionary conservation of the mutated positions, we added a collection of protein-oriented features that are based on functional aspects of the gene affected. We cross-validated DEOGEN on 36 825 polymorphisms, 20 821 deleterious SNVs, and 1038 INDELs from SwissProt. The multilevel contextualization of each (variant, protein) pair in DEOGEN provides a 10% improvement of MCC with respect to current state-of-the-art tools. AVAILABILITY AND IMPLEMENTATION: The software and the data presented here is publicly available at http://ibsquare.be/deogen CONTACT: : [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

The unraveling and control of protein stability at different temperatures is a fundamental problem in biophysics that is substantially far from being quantitatively and accurately solved, as it requires a precise knowledge of the temperature dependence of amino acid interactions. In this paper we attempt to gain insight into the thermal stability of proteins by designing a tool to predict the full stability curve as a function of the temperature for a set of 45 proteins belonging to 11 homologous families, given their sequence and structure, as well as the melting temperature (Tm) and the change in heat capacity (ΔCP) of proteins belonging to the same family. Stability curves constitute a fundamental instrument to analyze in detail the thermal stability and its relation to the thermodynamic stability, and to estimate the enthalpic and entropic contributions to the folding free energy. In summary, our approach for predicting the protein stability curves relies on temperature-dependent statistical potentials derived from three datasets of protein structures with targeted thermal stability properties. Using these potentials, the folding free energies (ΔG) at three different temperatures were computed for each protein. The Gibbs-Helmholtz equation was then used to predict the protein's stability curve as the curve that best fits these three points. The results are quite encouraging: the standard deviations between the experimental and predicted Tm's, ΔCP's and folding free energies at room temperature (ΔG25) are equal to 13° C, 1.3 kcal/(mol° C) and 4.1 kcal/mol, respectively, in cross-validation. The main sources of error and some further improvements and perspectives are briefly discussed.

BACKGROUND: The rational design of modified proteins with controlled stability is of extreme importance in a whole range of applications, notably in the biotechnological and environmental areas, where proteins are used for their catalytic or other functional activities. Future breakthroughs in medical research may also be expected from an improved understanding of the effect of naturally occurring disease-causing mutations on the molecular level. RESULTS: PoPMuSiC-2.1 is a web server that predicts the thermodynamic stability changes caused by single site mutations in proteins, using a linear combination of statistical potentials whose coefficients depend on the solvent accessibility of the mutated residue. PoPMuSiC presents good prediction performances (correlation coefficient of 0.8 between predicted and measured stability changes, in cross validation, after exclusion of 10% outliers). It is moreover very fast, allowing the prediction of the stability changes resulting from all possible mutations in a medium size protein in less than a minute. This unique functionality is user-friendly implemented in PoPMuSiC and is particularly easy to exploit. Another new functionality of our server concerns the estimation of the optimality of each amino acid in the sequence, with respect to the stability of the structure. It may be used to detect structural weaknesses, i.e. clusters of non-optimal residues, which represent particularly interesting sites for introducing targeted mutations. This sequence optimality data is also expected to have significant implications in the prediction and the analysis of particular structural or functional protein regions. To illustrate the interest of this new functionality, we apply it to a dataset of known catalytic sites, and show that a much larger than average concentration of structural weaknesses is detected, quantifying how these sites have been optimized for function rather than stability. CONCLUSION: The freely available PoPMuSiC-2.1 web server is highly useful for identifying very rapidly a list of possibly relevant mutations with the desired stability properties, on which subsequent experimental studies can be focused. It can also be used to detect sequence regions corresponding to structural weaknesses, which could be functionally important or structurally delicate regions, with obvious applications in rational protein design.

MOTIVATION: The rational design of proteins with modified properties, through amino acid substitutions, is of crucial importance in a large variety of applications. Given the huge number of possible substitutions, every protein engineering project would benefit strongly from the guidance of in silico methods able to predict rapidly, and with reasonable accuracy, the stability changes resulting from all possible mutations in a protein. RESULTS: We exploit newly developed statistical potentials, based on a formalism that highlights the coupling between four protein sequence and structure descriptors, and take into account the amino acid volume variation upon mutation. The stability change is expressed as a linear combination of these energy functions, whose proportionality coefficients vary with the solvent accessibility of the mutated residue and are identified with the help of a neural network. A correlation coefficient of R = 0.63 and a root mean square error of sigma(c) = 1.15 kcal/mol between measured and predicted stability changes are obtained upon cross-validation. These scores reach R = 0.79, and sigma(c) = 0.86 kcal/mol after exclusion of 10% outliers. The predictive power of our method is shown to be significantly higher than that of other programs described in the literature. AVAILABILITY: http://babylone.ulb.ac.be/popmusic

BACKGROUND: Superoxide dismutases (SODs) are ubiquitous metalloenzymes that play an important role in the defense of aerobic organisms against oxidative stress, by converting reactive oxygen species into nontoxic molecules. We focus here on the SOD family that uses Fe or Mn as cofactor. RESULTS: The SODa webtool http://babylone.ulb.ac.be/soda predicts if a target sequence corresponds to an Fe/Mn SOD. If so, it predicts the metal ion specificity (Fe, Mn or cambialistic) and the oligomerization mode (dimer or tetramer) of the target. In addition, SODa proposes a list of residue substitutions likely to improve the predicted preferences for the metal cofactor and oligomerization mode. The method is based on residue fingerprints, consisting of residues conserved in SOD sequences or typical of SOD subgroups, and of interaction fingerprints, containing residue pairs that are in contact in SOD structures. CONCLUSION: SODa is shown to outperform and to be more discriminative than traditional techniques based on pairwise sequence alignments. Moreover, the fact that it proposes selected mutations makes it a valuable tool for rational protein design.

UNLABELLED: Prelude&Fugue are bioinformatics tools aiming at predicting the local 3D structure of a protein from its amino acid sequence in terms of seven backbone torsion angle domains, using database-derived potentials. Prelude(&Fugue) computes all lowest free energy conformations of a protein or protein region, ranked by increasing energy, and possibly satisfying some interresidue distance constraints specified by the user. (Prelude&)Fugue detects sequence regions whose predicted structure is significantly preferred relative to other conformations in the absence of tertiary interactions. These programs can be used for predicting secondary structure, tertiary structure of short peptides, flickering early folding sequences and peptides that adopt a preferred conformation in solution. They can also be used for detecting structural weaknesses, i.e. sequence regions that are not optimal with respect to the tertiary fold. AVAILABILITY: http://babylone.ulb.ac.be/Prelude_and_Fugue.

Abstract Summary: PoPMuSiC is an efficient tool for rational computer-aided design of single-site mutations in proteins and peptides. Two types of queries can be submitted. The first option allows to estimate the changes in folding free energy for specific point mutations given by the user. In the second option, all possible point mutations in a given protein or protein region are performed and the most stabilizing or destabilizing mutations, or the neutral mutations with respect to thermodynamic stability, are selected. For each sequence position or secondary structure the deviation from the most stable sequence is moreover evaluated, which helps to identify the most suitable sites for the introduction of mutations. Availability: http://babylone.ulb.ac.be/popmusic Contact: [email protected] * To whom correspondence should be addressed.