Residue conservation and solvent accessibility are (almost) all you need for predicting mutational effects in proteins

Residue conservation and solvent accessibility are (almost) all you need for predicting mutational effects in proteins | bioRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-M677548'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search New Results Residue conservation and solvent accessibility are (almost) all you need for predicting mutational effects in proteins Matsvei Tsishyn , Pauline Hermans , Fabrizio Pucci , Marianne Rooman doi: https://doi.org/10.1101/2025.02.03.636212 Matsvei Tsishyn 1 Computational Biology and Bioinformatics, Université Libre de Bruxelles , 1050 Brussels, Belgium 2 Interuniversity Institute of Bioinformatics in Brussels , 1050 Bruxelles, Belgium Find this author on Google Scholar Find this author on PubMed Search for this author on this site Pauline Hermans 1 Computational Biology and Bioinformatics, Université Libre de Bruxelles , 1050 Brussels, Belgium 2 Interuniversity Institute of Bioinformatics in Brussels , 1050 Bruxelles, Belgium Find this author on Google Scholar Find this author on PubMed Search for this author on this site Fabrizio Pucci 1 Computational Biology and Bioinformatics, Université Libre de Bruxelles , 1050 Brussels, Belgium 2 Interuniversity Institute of Bioinformatics in Brussels , 1050 Bruxelles, Belgium Find this author on Google Scholar Find this author on PubMed Search for this author on this site For correspondence: fabriziopucci81{at}gmail.com Marianne Rooman 1 Computational Biology and Bioinformatics, Université Libre de Bruxelles , 1050 Brussels, Belgium 2 Interuniversity Institute of Bioinformatics in Brussels , 1050 Bruxelles, Belgium Find this author on Google Scholar Find this author on PubMed Search for this author on this site Abstract Full Text Info/History Metrics Supplementary material Preview PDF Abstract 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 (LOR) of wild-type and mutated residue frequencies in evolutionary related proteins, when scaled by the residue’s relative solvent accessibility (RSA), 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 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 . Contact matsvei.tsishyn{at}ulb.be , fabrizio.pucci{at}ulb.be Introduction Accurately estimating the fitness of variants is essential both from a biomedical perspective, to deepen our understanding of the mechanisms underlying pathogenesis [ 1 ], and from a biotech-nological perspective, to improve protein engineering approaches [ 2 ]. As a result, an impressive number of computational tools have been developed over the last decade to predict the effects of variants on different protein biophysical properties [ 3 , 4 , 5 ]. These tools are characterized by a wide range of architectures, ranging from simple linear models applied to a few features to complex deep learning methods. In recent years, the field has witnessed an even more remarkable growth, with the emergence of protein language models (pLMs), which have significantly advanced variant effect prediction [ 5 , 6 ]. However, these deep learning techniques also come with notable drawbacks. Their immense number of parameters requires extensive training, making them computationally expensive and more prone to overfitting. While there are methods to mitigate overfitting, it remains challenging to disentangle true biophysical properties from unwanted biases and to achieve good generalizability. Additionally, their inherent complexity often makes it nearly impossible to extract meaningful biophysical insights from their predictions. An alternative approach is to introduce simple prediction models with biological significance that retain the same accuracy as these more complex approaches. Following this direction [ 7 ], we recently showed how a simple evolutionary score, when scaled by the relative solvent accessibility (RSA) of the mutated residues, can accurately predict changes in folding free energy upon mutations. The model is extremely simple, interpretable and performant, and has no free parameters to optimize. Its score reflects the impact of a variant on protein fitness, which is broadly defined as the ability of the protein to effectively perform its biological function. As such, it is related not only to stability but also to other protein properties. We thus extended the analysis of this model, called RSALOR, by providing additional evidence of its accuracy across a broader range of protein biophysical properties, including stability, binding affinity, organismal fitness, activity, and expression, using ProteinGym [ 6 ], a widely used benchmark dataset specifically designed for this purpose. A graphical representation of the model is provided in Fig. 1 . Download figure Open in new tab Figure 1. Graphical representation of the RSALOR model. RSALOR model We present a very simple, independent-site, unsupervised approach for mutational effects prediction that combines evolutionary and structural information. In this section, we outline the key steps of the RSALOR model. Detailed implementation aspects are described in Supplementary Materials (SM) Sections 1 and 2. The evolutionary information used in the model is derived from amino acid frequencies at the mutated position in a multiple sequence alignment (MSA), using the LOR between wild-type and mutant amino acid frequencies. The MSA is first curated by removing redundant identical sequences and those falling within the “twilight zone” [ 8 ] (i.e., sequences too evolutionary distant from the target sequence based on a sequence identity criterion). Indeed, we observed that the presence of very distant sequences adds noise rather than improving RSALOR predictions. Since MSAs can be dominated by clusters of closely related sequences, we computed the “weighted” amino acid frequencies by reducing the contribution of sequences from larger clusters, as done in coevolutionary models, e.g. [ 9 , 10 ]. We discussed and analyzed the impact of the weighting step in SM Section 3.2, showing that it mildly but consistently improves our model’s performance. To prevent LOR values from diverging and to handle the lack of information in small MSAs, we applied regularization to these frequencies. Using the weighted and regularized frequencies f i ( wt ) and f i ( mt ) for the wild-type and mutant amino acids at position i , the LOR is defined as: The sign of LOR is defined such that the result of mutations from a highly represented amino acid wt to a less represented amino acid mt is positive, which generally corresponds to a decrease in protein stability or fitness. As structural information, we used the per-residue RSA, reflecting the observation that mutations in the protein core tend to have a greater impact than those on the surface. The simple product of LOR and the “complement” of RSA defines the RSALOR: Since the RSA factor in this equation is always positive, the sign of RSALOR is the same as the sign of LOR. Note that RSALOR nearly perfectly preserves the symmetry property (i.e., the effect of a mutation from wt to mt is opposite to the effect of the mutation from mt to wt ). Indeed, while the evolutionary component LOR is perfectly symmetric, the fact that RSA is calculated using only the wild-type structure can, in principle, introduce asymmetry into the model. However, as shown in SM Section 3.4, this approximation does not substantially impact the predictions, which remain almost perfectly symmetric. This symmetry, which is often violated by predictors, has been shown to be an important feature in the prediction of changes in protein stability and binding affinity [ 11 , 12 , 13 ]. RSALOR implementation We provide RSALOR as a freely available, easy-to-install, and user-friendly Python package. It can be installed by cloning our GitHub repository at github.com/3BioCompBio/RSALOR or via the Python Package Index (PyPI) using pip . It takes as input the MSA of the target protein and its three-dimensional structure in PDB format. The package automatically maps RSA values extracted from the structure to the corresponding positions in the MSA, even if the template structure contains missing residues or is homologous but not identical to the target sequence of the MSA. Indeed, RSA values are relatively robust to small structural changes. The model’s performance therefore remains almost unchanged when using structures with a few amino acid substitutions (see SM Section 3.5 for details). The RSALOR package outputs or saves to a CSV file the following information for each possible single-site mutation in the target protein: the frequencies of gaps, wild-type and mutant residues in the MSA; the RSA of the mutated residue; and the LOR and RSALOR scores of the mutation. RSALOR performances In [ 7 ], we evaluated RSALOR on its ability to predict the impact of mutations on protein stability. To further assess the robustness of the model, we tested it on ProteinGym [ 6 ] consisting of 218 standardized deep mutational scanning (DMS) experiments, covering a total of million mutations with annotated experimental effects on protein stability, binding affinity, fitness, activity, and expression. View this table: View inline View popup Download powerpoint Table 1. Average per-DMS Spearman correlations across ProteinGym subclasses (categorized by DMS target properties), comparing the RSALOR model with some of the top-performing models from the ProteinGym benchmark. Only single-site mutations were considered. Note that ProteinGym’s benchmark uses a slightly different method of averaging correlations, so their values and ours do not always perfectly match. * STR = structure-based, ALI = alignment-based To ensure a fair comparison with the other benchmarked models, we used the MSAs and structures provided by the ProteinGym repository without any modifications. These structures are AlphaFold-generated models [ 22 ], as the target sequences of most DMS experiments are not, or only partially, covered by experimental structures. Importantly, this means that our model does not rely on the availability of high-quality experimental structures. We additionally assessed the robustness of our predictions using alternative MSAs and structures, and observed essentially the same results (see details in SM Section 3.5). While we present here a summary of the results, a more comprehensive analysis is provided in SM Section 3.1. It includes both overall performances and per-category performances on single-site and all (single-site and multiple) mutations from ProteinGym, evaluated using various metrics and compared among 27 different predictors (including 19 pLM-based models). We first focused on the 700, 000 single-site mutations in the dataset. In Tab. 1 , we present a comparison of the Spearman correlations, averaged over all DMS experiments or over specific DMS categories. The predictions of RSALOR were compared with some of the top-performing tools included in the ProteinGym benchmark. Our results show that the simple RSALOR model achieves performance in line with the other state-of-the-art methods, with an average Spearman correlation of 0.473 across all 218 datasets. Among the 27 models in the unsupervised category, only ProSST [ 14 ], which combines structure and pLM features, achieves better results. We note that the structural contribution, RSA, has a variable impact on performance depending on the DMS target property. Although it provides great improvements to the LOR score on stability and binding datasets, it enhances accuracy to a lesser extent for expression, activity, and fitness datasets. This is consistent considering that stability and binding affinity are more directly related to protein structure. For instance, the RSA score alone outperforms most benchmarked predictions on stability datasets. In addition, we have shown that using RSA computed from more appropriate input 3D conformations can further boost predictions (see SM Section 3.6). For example, when studying the impact of mutations on protein–protein binding affinity, using the structure of the protein complex instead of the monomeric structure provided by the ProteinGym dataset leads to substantially improved performance. It is worth noting, as already highlighted in the ProteinGym benchmark [ 6 ], that predictions also vary greatly between datasets, with some DMSs being exceptionally well predicted (Spearman correlation above 0.7), while in others, all methods essentially fail. In contrast, the number of homologous sequences in the input MSA has only a minor impact on the performance of RSALOR (see details in SM Section 3.3). To predict the effect of multiple mutations using the RSALOR model, we made the approximation that there are no epistatic effects. Therefore, the effect of a multiple mutation is simply the sum of the effects of its individual single-site mutations. Even with this simplistic assumption, RSALOR achieved a correlation of 0.484 on all ProteinGym mutations, outperformed only by ProSST [ 14 ] and PoET [ 15 ] (with correlations of 0.523 and 0.490, respectively). All values are provided in SM Section 3.1. We would like to underline that the RSALOR model is clearly a rough approximation for estimating the effects of protein mutations. First, it assumes that mutations at fully exposed residues (with a RSA of 100%) have no effect on the protein. While the link between the RSA of mutated residues and mutational impacts is well known [ 23 , 24 ], the strength of its effect can vary significantly depending on the biophysical property considered. Second, RSALOR completely ignores epistatic effects and potential evolutionary information from other residue positions. While their contribution could be less significant in describing protein stability [ 7 , 25 ], they seem to play an essential roles in protein activity and fitness [ 25 , 26 ]. Despite these limitations, the model is on par with or slightly outperforms nearly all benchmarked models, highlighting that predicting mutational landscapes remains a challenge for current state-of-the-art methods. Remarkably, combining the predictions of other models with RSA values (using Eq. 2 ) substantially improves the performance for almost all of the 27 benchmarked predictors. This holds true even for models that already incorporate structural information as input (see details in SM Section 4). We thus show that effectively incorporating RSA and other types of structural knowledge into evolutionary-or pLM-based models can lead to improved performance. Finally, an important feature of RSALOR is its ease of use. It requires no model training or external dependencies, is easily installed with a single pip command, and runs in a straight-forward manner (see the GitHub repository). From a computational perspective, RSALOR is highly efficient. Its weighting step, being the most computationally intensive, is implemented in C++ and supports multi-threading. For instance, we evaluated the 2.5 millions mutations from ProteinGym in less than 20 minutes on a laptop using 8 CPUs. Results for each individual protein were generated in a time range of 1 second to 1 minute. Acknowledgments We acknowledge financial support from the Belgian Fund for Scientific Research (F.R.S.-FNRS) through a PDR project. M.T. benefits from a FNRS-FRIA PhD grant. P.H. benefits from a Win4Doc grant from SPW Recherche of the Walloon Region. Funder Information Declared Fonds de la recherche scientifique, FNRS, Belgium Service Public de Wallonie, Belgium Footnotes Improved version; further analysis performed; typos corrected. References [1]. ↵ Douglas M Fowler , David J Adams , Anna L Gloyn , William C Hahn , Debora S Marks , Lara A Muffley , James T Neal , Frederick P Roth , Alan F Rubin , Lea M Starita , et al. An atlas of variant effects to understand the genome at nucleotide resolution . Genome Biology , 24 ( 1 ): 147 , 2023 . OpenUrl CrossRef PubMed [2]. ↵ Chase R Freschlin , Sarah A Fahlberg , and Philip A Romero . Machine learning to navigate fitness landscapes for protein engineering . Current opinion in biotechnology , 75 : 102713 , 2022 . OpenUrl CrossRef PubMed [3]. ↵ Fabrizio Pucci , Martin Schwersensky , and Marianne Rooman . Artificial intelligence challenges for predicting the impact of mutations on protein stability . Current opinion in structural biology , 72 : 161 – 168 , 2022 . OpenUrl CrossRef PubMed [4]. ↵ Benjamin J Livesey and Joseph A Marsh . Updated benchmarking of variant effect predictors using deep mutational scanning . Molecular systems biology , 19 ( 8 ): e11474 , 2023 . OpenUrl CrossRef PubMed [5]. ↵ Ruchir Rastogi , Ryan Chung , Sindy Li , Chang Li , Kyoungyeul Lee , Junwoo Woo , DongWook Kim , Changwon Keum , Giulia Babbi , Pier Luigi Martelli , et al. Critical assessment of missense variant effect predictors on disease-relevant variant data . bioRxiv , pages 2024 – 06 , 2024 . [6]. ↵ Pascal Notin , Aaron Kollasch , Daniel Ritter , Lood Van Niekerk , Steffanie Paul , Han Spinner , Nathan Rollins , Ada Shaw , Rose Orenbuch , Ruben Weitzman , et al. ProteinGym: Large-scale benchmarks for protein fitness prediction and design . Advances in Neural Information Processing Systems, 36 , 2024 . [7]. ↵ Pauline Hermans , Matsvei Tsishyn , Martin Schwersensky , Marianne Rooman , and Fabrizio Pucci . Exploring evolution to uncover insights into protein mutational stability . Molecular Biology and Evolution , page msae267, 2024 . [8]. ↵ Burkhard Rost . Twilight zone of protein sequence alignments . Protein engineering , 12 ( 2 ): 85 – 94 , 1999 . OpenUrl CrossRef PubMed Web of Science [9]. ↵ Martin Weigt , Robert A White , Hendrik Szurmant , James A Hoch , and Terence Hwa . Identification of direct residue contacts in protein–protein interaction by message passing . Proceedings of the National Academy of Sciences , 106 ( 1 ): 67 – 72 , 2009 . OpenUrl Abstract / FREE Full Text [10]. ↵ Faruck Morcos , Andrea Pagnani , Bryan Lunt , Arianna Bertolino , Debora S Marks , Chris Sander , Riccardo Zecchina , José N Onuchic , Terence Hwa , and Martin Weigt . Direct-coupling analysis of residue coevolution captures native contacts across many protein families . Proceedings of the National Academy of Sciences , 108 ( 49 ): E1293 – E1301 , 2011 . OpenUrl Abstract / FREE Full Text [11]. ↵ Fabrizio Pucci , Katrien V Bernaerts , Jean Marc Kwasigroch , and Marianne Rooman . Quantification of biases in predictions of protein stability changes upon mutations . Bioinformatics , 34 ( 21 ): 3659 – 3665 , 2018 . OpenUrl CrossRef PubMed [12]. ↵ Dinara R Usmanova , Natalya S Bogatyreva , Joan Ariño Bernad , Aleksandra A Eremina , Anastasiya A Gorshkova , German M Kanevskiy , Lyubov R Lonishin , Alexander V Meister , Alisa G Yakupova , Fyodor A Kondrashov , et al. Self-consistency test reveals systematic bias in programs for prediction change of stability upon mutation . Bioinformatics , 34 ( 21 ): 3653 – 3658 , 2018 . OpenUrl CrossRef PubMed [13]. ↵ Matsvei Tsishyn , Fabrizio Pucci , and Marianne Rooman . Quantification of biases in predictions of protein–protein binding affinity changes upon mutations . Briefings in bioinformatics , 25 ( 1 ): bbad491 , 2024 . OpenUrl CrossRef [14]. ↵ Mingchen Li , Yang Tan , Xinzhu Ma , Bozitao Zhong , Huiqun Yu , Ziyi Zhou , Wanli Ouyang , Bingxin Zhou , Pan Tan , and Liang Hong . ProSST: Protein language modeling with quantized structure and disentangled attention . In The Thirty-eighth Annual Conference on Neural Information Processing Systems , 2024 . [15]. ↵ Timothy Truong Jr and Tristan Bepler . Poet: A generative model of protein families as sequences-of-sequences . Advances in Neural Information Processing Systems , 36 : 77379 – 77415 , 2023 . OpenUrl [16]. Jin Su , Chenchen Han , Yuyang Zhou , Junjie Shan , Xibin Zhou , and Fajie Yuan . SaProt: Protein language modeling with structure-aware vocabulary . bioRxiv , pages 2023 – 10 , 2023 . [17]. Céline Marquet , Julius Schlensok , Marina Abakarova , Burkhard Rost , and Elodie Laine . Expert-guided protein language models enable accurate and blazingly fast fitness prediction . Bioinformatics , 40 ( 11 ): btae621 , 2024 . OpenUrl CrossRef PubMed [18]. Pascal Notin , Lood Van Niekerk , Aaron W Kollasch , Daniel Ritter , Yarin Gal , and Debora S Marks . TranceptEVE: Combining family-specific and family-agnostic models of protein sequences for improved fitness prediction . bioRxiv , pages 2022 – 12 , 2022 . [19]. Elodie Laine , Yasaman Karami , and Alessandra Carbone . GEMME: a simple and fast global epistatic model predicting mutational effects . Molecular biology and evolution , 36 ( 11 ): 2604 – 2619 , 2019 . OpenUrl CrossRef PubMed [20]. Jonathan Frazer , Pascal Notin , Mafalda Dias , Aidan Gomez , Joseph K Min , Kelly Brock , Yarin Gal , and Debora S Marks . Disease variant prediction with deep generative models of evolutionary data . Nature , 599 ( 7883 ): 91 – 95 , 2021 . OpenUrl CrossRef PubMed [21]. Zeming Lin , Halil Akin , Roshan Rao , Brian Hie , Zhongkai Zhu , Wenting Lu , Nikita Smetanin , Robert Verkuil , Ori Kabeli , Yaniv Shmueli , et al. Evolutionary-scale prediction of atomic-level protein structure with a language model . Science , 379 ( 6637 ): 1123 – 1130 , 2023 . OpenUrl CrossRef PubMed [22]. ↵ John Jumper , Richard Evans , Alexander Pritzel , Tim Green , Michael Figurnov , Olaf Ronneberger , Kathryn Tunyasuvunakool , Russ Bates , Augustin Žídek , Anna Potapenko , et al. Highly accurate protein structure prediction with alphafold . nature , 596 ( 7873 ): 583 – 589 , 2021 . OpenUrl CrossRef PubMed [23]. ↵ Qiong Wei , Qifang Xu , and Roland L Dunbrack Jr . Prediction of phenotypes of missense mutations in human proteins from biological assemblies . Proteins: Structure, Function, and Bioinformatics , 81 ( 2 ): 199 – 213 , 2013 . OpenUrl CrossRef [24]. ↵ François Ancien , Fabrizio Pucci , Maxime Godfroid , and Marianne Rooman . Prediction and interpretation of deleterious coding variants in terms of protein structural stability . Scientific reports , 8 ( 1 ): 4480 , 2018 . OpenUrl CrossRef PubMed [25]. ↵ Matt Sternke , Katherine W Tripp , and Doug Barrick . Protein stability is determined by single-site bias rather than pairwise covariance . bioRxiv , pages 2025 – 01 , 2025 . [26]. ↵ William P Russ , Matteo Figliuzzi , Christian Stocker , Pierre Barrat-Charlaix , Michael Socolich , Peter Kast , Donald Hilvert , Remi Monasson , Simona Cocco , Martin Weigt , et al. An evolution-based model for designing chorismate mutase enzymes . Science , 369 ( 6502 ): 440 – 445 , 2020 . OpenUrl Abstract / FREE Full Text View the discussion thread. Back to top Previous Next Posted May 25, 2025. Download PDF Supplementary Material Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Residue conservation and solvent accessibility are (almost) all you need for predicting mutational effects in proteins Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Residue conservation and solvent accessibility are (almost) all you need for predicting mutational effects in proteins Matsvei Tsishyn , Pauline Hermans , Fabrizio Pucci , Marianne Rooman bioRxiv 2025.02.03.636212; doi: https://doi.org/10.1101/2025.02.03.636212 Share This Article: Copy Citation Tools Residue conservation and solvent accessibility are (almost) all you need for predicting mutational effects in proteins Matsvei Tsishyn , Pauline Hermans , Fabrizio Pucci , Marianne Rooman bioRxiv 2025.02.03.636212; doi: https://doi.org/10.1101/2025.02.03.636212 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Bioinformatics Subject Areas All Articles Animal Behavior and Cognition (7635) Biochemistry (17697) Bioengineering (13895) Bioinformatics (41951) Biophysics (21456) Cancer Biology (18594) Cell Biology (25520) Clinical Trials (138) Developmental Biology (13381) Ecology (19903) Epidemiology (2067) Evolutionary Biology (24323) Genetics (15612) Genomics (22510) Immunology (17737) Microbiology (40401) Molecular Biology (17183) Neuroscience (88622) Paleontology (667) Pathology (2833) Pharmacology and Toxicology (4825) Physiology (7644) Plant Biology (15158) Scientific Communication and Education (2046) Synthetic Biology (4296) Systems Biology (9825) Zoology (2271)