Co-translational assembly promotes functional diversification of paralogous proteins

Co-translational assembly promotes functional diversification of paralogous 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 Co-translational assembly promotes functional diversification of paralogous proteins View ORCID Profile Saurav Mallik , View ORCID Profile Angel F. Cisneros , View ORCID Profile Christian R. Landry , View ORCID Profile Emmanuel D. Levy doi: https://doi.org/10.1101/2025.01.22.634331 Saurav Mallik 1 Department of Molecular Genetics, Weizmann Institute of Science , Rehovot 7600001, Israel 2 Department of Chemical and Structural Biology, Weizmann Institute of Science , Rehovot 7600001, Israel Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Saurav Mallik For correspondence: saurav.mallik{at}weizmann.ac.il christian.landry{at}bio.ulaval.ca emmanuel.Levy{at}weizmann.ac.il Angel F. Cisneros 1 Department of Molecular Genetics, Weizmann Institute of Science , Rehovot 7600001, Israel 3 Département de biochimie, de microbiologie et de bio-informatique, Faculté des sciences et de génie, Université Laval G1V 0A6 , Québec, Canada 4 Institut de biologie intégrative et des systèmes, Université Laval G1V 0A6 , Québec, Canada 5 PROTEO, Le regroupement québécois de recherche sur la fonction, l’ingénierie et les applications des protéines, Université Laval G1V 0A6 , Québec, Canada 6 Centre de recherche sur les données massives, Université Laval G1V 0A6 , Québec, Canada Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Angel F. Cisneros Christian R. Landry 3 Département de biochimie, de microbiologie et de bio-informatique, Faculté des sciences et de génie, Université Laval G1V 0A6 , Québec, Canada 4 Institut de biologie intégrative et des systèmes, Université Laval G1V 0A6 , Québec, Canada 5 PROTEO, Le regroupement québécois de recherche sur la fonction, l’ingénierie et les applications des protéines, Université Laval G1V 0A6 , Québec, Canada 6 Centre de recherche sur les données massives, Université Laval G1V 0A6 , Québec, Canada Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Christian R. Landry For correspondence: saurav.mallik{at}weizmann.ac.il christian.landry{at}bio.ulaval.ca emmanuel.Levy{at}weizmann.ac.il Emmanuel D. Levy 1 Department of Molecular Genetics, Weizmann Institute of Science , Rehovot 7600001, Israel 2 Department of Chemical and Structural Biology, Weizmann Institute of Science , Rehovot 7600001, Israel 7 Department of Molecular and Cellular Biology, University of Geneva , 30, quai Ernest-Ansermet CH-1211, Geneva 4, Switzerland Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Emmanuel D. Levy For correspondence: saurav.mallik{at}weizmann.ac.il christian.landry{at}bio.ulaval.ca emmanuel.Levy{at}weizmann.ac.il Abstract Full Text Info/History Metrics Supplementary material Data/Code Preview PDF ABSTRACT Homomeric proteins are ubiquitous and mediate myriads of cellular functions. When a gene encoding a homomer duplicates, the resulting paralogs can either form distinct homomers, or evolve into a heteromer containing both paralogs. While such events have extensively shaped proteomes, the molecular mechanisms driving these fates and their associated functional consequences remain largely unknown. Here, we conducted a comprehensive phylogenomic analysis tracing gene duplication histories of 7,377 human paralogs across the eukaryotic lineage and identified their fates using protein interaction data. Simulations and data analyses show that cellular constraints must act as barriers to disfavor heteromerization and promote homomerization. We found that multiple cellular and molecular constraints can serve as barriers, including the lack of co-expression and co-localization. The main barrier, however, is co-translational assembly, which naturally promotes the self-assembly of each paralog from its corresponding mRNA, thus hindering heteromerization. We further established that heteromerization constrains functional divergence, with homomeric paralogs exhibiting stronger signatures of adaptive evolution and functional divergence compared to heteromeric paralogs. Together, these findings identify key biochemical and cellular properties that explain protein function diversification following gene duplication. One Sentence Summary Co-translational assembly drives the selective homo-oligomerization of paralogs, which in turn promotes their functional divergence. Download figure Open in new tab INTRODUCTION Gene duplication occurs frequently in eukaryotic genomes, arising through processes that range from the duplication of short chromosomal segments to that of entire chromosomes or genomes 1 – 5 . In principle, since the two duplicated copies (paralogs) are initially functionally redundant, one copy generally loses its function by accumulating mutations ( pseudogenization ) 6 – 9 . Copies that survive may either retain their ancestral function ( conservation ) or diverge by splitting the ancestral function ( subfunctionalization ) or by acquiring new functions ( neofunctionalization ) 10 – 17 . Gene duplication is therefore a major source of functional innovation, which has prompted a large body of research to explore and rationalize the diverse evolutionary trajectories of paralogous pairs. These studies include the mathematical modeling of mutational and selection forces underlying paralog divergence 4 , 10 , 17 – 20 , the experimental characterization of paralogs’ functional redundancies 21 – 26 , or computational approaches characterizing their evolution and divergence in sequence and function 27 – 31 . Although functional divergence appears to be key for the long term maintenance of paralogous pairs, many remain functionally redundant 17 , 32 , 33 , which has important implications in cellular networks 34 , development 35 – 39 , immune system 40 , neuronal regeneration 41 , and cell cycle 42 , 43 . For example, Tenascin is important for tissue remodeling and collagen fibrillogenesis; yet, its knockout does not lead to severe phenotypic defects in mice unless combined with knockouts of its paralogs 37 , 41 . Similarly, knockout of protein tyrosine kinase Src does not block fibroblast cells’ G2 to mitosis transition, unless combined with the knockout of its paralogs Fyn and Yes1 44 . Ultimately, the degree of functional divergence among paralogous pairs spans a wide spectrum, and general principles explaining why certain pairs diverge more than others remain elusive. The evolution of a protein’s function can be constrained by multiple factors. For example, two proteins can become functionally coupled by direct physical interactions or via links in metabolic, regulatory, or signaling pathways 45 . These couplings can restrict the evolutionary trajectories of paralogs, limiting the paths they would have otherwise taken if evolving independently 45 – 49 . In particular, the dependency between subunits within a macromolecular complex can result in them being functional only in the oligomeric state. For example, this is seen in oligomeric enzymes in which the catalytic sites are shared between subunits 50 , and also in large complexes such as the eukaryotic proteasome complex, where 14 paralogous subunits form the functional 20S core and are all essential for function 51 . Selection for new functions during neofunctionalization and the loss of ancestral ones during subfunctionalization are key factors driving the maintenance of gene duplicates. Here, we asked whether structural and biophysical properties of proteins could influence these fates. We focused on the physical association between paralogs, which is systematically inherited from the duplication of homomer-encoding genes. Homomer-forming proteins are simple model systems for this study, because their duplications naturally introduce paralogs with or without direct cross-paralog interactions 52 , 53 . In the absence of any constraints, paralogs originating from the gene duplication of a homomeric protein would form a mixture of homo- and heteromeric complexes 49 , 54 ( Mixed scenario ). Upon subsequent sequence divergence, this Mixed state might be retained. Alternatively, the two copies might lose their heteromeric interaction, resulting in two independent homomers ( Homomeric scenario ); or, they might lose the capacity to self-interact, resulting in a single heteromeric complex ( Heteromeric scenario ) 55 – 59 . These alternative fates have been studied extensively, by experimental 58 – 61 or phylogenetic characterization of specific complexes 62 – 65 , or by genomic analyses examining their relative abundances 54 – 57 . Yet, the underlying molecular determinants of these fates and whether they associate with specific functional outcomes, remain largely unknown. To address these questions, we conducted a comprehensive phylogenomic analysis tracing the gene duplication histories of human paralogous proteins across the eukaryotic lineage and examined their oligomeric fates using protein interaction data. This analysis along with simulations show that heteromerization is favored by neutral drift, unless constraints in the cellular environment act as barriers to the heteromeric interaction. We found that multiple cellular and molecular constraints can act as such barriers, including changes in gene expression profile, protein subcellular localization (spatiotemporal separation), or modifications at the oligomeric interfaces (structural incompatibility). Notably, co-translational assembly naturally induces each paralog to self-assemble from its respective mRNA and appeared as the barrier hindering paralog heteromerization the most. We also showed that direct physical interaction restricts the functional divergence of paralogous pairs. Such heteromeric paralogs tend to evolve in the nearly-neutral regime, accumulate substitutions and insertion/deletions more slowly than homomeric paralogs, and remain functionally more similar. By contrast, homomeric paralogs tend to evolve in the adaptive regime, accumulate evolutionary changes more rapidly, and increasingly diverge in structure and function with duplication age. Interestingly, the onset of their functional divergence appears to follow the apparition of the barrier to heteromerization. Together, these findings unveil a fundamental connection between co-translational assembly and paralog heteromerization, highlighting its widespread impact on the evolutionary origin of new protein functions. RESULTS Tracing the duplications of human paralogous genes across the eukaryotic tree of life We traced gene duplication histories in the eukaryotic lineage leading to humans ( Methods ). To this end, we collected proteome sequences for 106 taxa representing all major eukaryotic clades; with an additional 11 Proteobacteria and six Asgard archaea proteomes that served as the outgroups ( Table S1 ). We obtained the evolutionary relationships among these 123 organisms from the TimeTree database 66 and constructed an eukaryotic Tree of Life (which serves as the species tree, Figure 1A ). We also adhered to the branch lengths obtained from TimeTree, which are species divergence times in million years, derived from molecular dating calibrated with geological records 66 . Download figure Open in new tab Figure 1. Inferring duplications of human paralogous genes across the eukaryotic tree of life. ( A ) The eukaryotic Tree of Life represents the phylogenetic relationships (obtained from the TimeTree database 66 ) among the 123 organisms (106 eukaryotes, 11 Proteobacteria, and 6 Asgard archaea) with fully sequenced genomes. Closely related organisms were grouped to represent the major phylogenetic clades and the #taxa are mentioned ( e . g ., Xenopus laevis and Geotrypetes seraphini were grouped into Amphibia ). Leaves represent extant clades, while internal nodes (A1, A2, … A21) represent their common ancestors. Branch lengths are in million years ago ( mya ) unit, and reflect data from TimeTree. Organisms were sketched by the authors. The Last Eukaryotic Common Ancestor, or LECA (A21) represents the root of the eukaryotic sub-tree. Gene duplication events of human paralogs were phylogenetically inferred across the eukaryotic lineage. Vertical bars represent the frequency of inferred high- confidence duplication events at each ancestral node. Those giving rise to Homomeric, Mixed , and Heteromeric pairs in human cells are highlighted in orange, yellow, and green. Duplications for which the interaction status could not be reliably assigned are in dark blue. ( B ) Human protein-coding genes are classified into various functional categories according to KEGG annotations 95 . For each functional category, the total number of annotated genes and the number of homomer-derived paralogous genes are shown (only those categories with ≥50 genes are plotted). ( C ) For each functional category, the relative proportion of gene duplications (normalized by the row-sum) across the ancestral nodes is shown as a heatmap. ( D ) Same as C , for the percentage of duplications that gave rise to a pair of heteromeric paralogs. To detect human paralogous genes, we compared more than 2.1 million proteins (each corresponding to a gene) using all- versus -all BLAST 67 and then clustered them using a Markov cluster algorithm 68 , 69 into 60259 protein families. Out of those, 5330 families included ≥2 paralogous human genes. To identify paralogous pairs that originated from homomeric ancestors, we examined their current oligomeric states. Molecular interaction data for human proteins were obtained from publicly available databases and include (i) crystallography, electron microscopy 70 , and AlphaFold2-modeled structural data on protein complexes 71 , (ii) subunit composition data for curated macromolecular complexes 72 , 73 , and (iii) experimental protein-protein interaction data 74 – 80 . These different datasets were integrated and filtered to minimize false positives, which yielded a high-quality set of 4,940 homomeric proteins and 142,226 heteromeric pairs ( Methods, Data S1 ). Each paralogous pair was then classified into one of the three categories: Homomeric, Heteromeric , and Mixed . Paralogs featuring these oligomeric states are parsimoniously expected to have originated from ancestral homomers 54 – 57 . We identified 1998 families including 7377 such paralogs ( Data S2 ), highlighting that ∼35% (over a third) of the human proteome originates from ancestral homomers. We used two complementary phylogenetic approaches to delineate the timing of gene duplication events ( Methods ). In the first approach, we inferred gene trees for each family and compared with the species tree to date the duplication events. In the second approach, given the gene copy numbers (GCN) for each species, we inferred duplication events for each family by a probabilistic reconstruction of the ancestral GCNs 81 . Out of the 5381 inferred duplications, 4199 were considered high-confidence, as both approaches assigned the event at the same branch with ≥70 bootstrap support or ≥0.7 posterior probability ( Data S2 ). Our results showed that the human paralogous pairs duplicated at various stages of eukaryotic evolution ( Figure 1A ). About 33% of the duplications occurred in a single node Gnathostomata (A12, 1439 duplications), which coincides with the two/three rounds of whole-genome duplications during the origin of vertebrates 1 , 82 – 85 . An additional ∼10% and ∼8% of the duplications occurred in Metazoa and LECA (A17 and A21, 431 and 355 duplications). Indeed, rampant gene duplications preceded the origin of eukaryotes 86 – 91 and metazoans 92 – 94 , and the functional diversifications of the resulting paralogs are believed to have driven the complexification of eukaryotic cells and multicellular metazoan life 86 , 92 . What were the oligomeric fates of these duplicates? Since 1168 out of the 1998 families harbor nested duplications, the oligomeric fate of each duplication was inferred by parsimony ( Methods ). Results show that out of 4199 high-confidence duplications, 596, 640, and 983 gave rise to Heteromeric, Mixed , and Homomeric pairs respectively ( Figure 1A , Data S2 ). The fate of 1980 duplications remained inconclusive due to the absence of experimental data on their physical interactions, and these were excluded from further analysis. Thus, ∼55% of the characterized high-confidence duplications resulted in the emergence of heteromeric paralogs ( Heteromeric + Mixed ), whereas the remaining ∼45% retained the ancestral homomeric state. Next, we examined how these duplications contributed to the functional repertoire of mammalian proteomes. The human proteome was classified into different functional categories according to the KEGG database 95 annotations ( Figure 1B , Methods ). Homomer-derived paralogs contribute the least (∼21%) to functions related to translation machinery (transfer RNA biogenesis, ribosome biogenesis, and ribosome) and the most (∼70%) to lectins, nuclear receptors, proteoglycans, and kinases. Consistent with Figure 1A , three episodic bursts of duplications ( Gnathostomata, Metazoa , and LECA) contributed to the expansion of most of these functional categories ( Figure 1C ). We asked whether homo-versus heteromeric paralogs were associated with specific episodes of gene duplications. To that end, the fraction of duplications that gave rise to heteromeric pairs was computed for each branch of the species tree. However, no such coincidences were found and the fraction of heteromeric paralogs did not correlate with the frequency of duplication across the ancestral branches (Pearson R = 0.14, Figure S1 ). Overall, heteromeric fates dominate (>80%) in the following functional classes: DNA replication, repair, recombination, transcription, splicing, and proteasome, but are rare (<30%) in membrane proteins, chaperones, and glycosyltransferases ( Figure 1D ). Taken together, these results highlight that over a third of the human proteome has descended from gene duplications of ancestral homomers. About 55% of the characterized duplication events gave rise to heteromeric paralogs, which are particularly frequent among functions associated with the central dogma. Barriers against paralog heteromerization drive the evolution of homomeric paralogs What are the molecular determinants of different oligomeric fates? To address this question, we previously developed a biophysical model simulating the neutral accumulation of mutations in two duplicates and their impact on the probability that paralogs are maintained in homomeric vs . heteromeric states 96 ( Figure 2A ). These simulations were based on two important assumptions. First, that oligomerization occurs by random encounters of translated and folded monomers. Second, that mutations are fixed under a neutral model with respect to the three molecular complexes (AA, AB, and BB); specifically, (i) selection acts only on the total number of complexes without favoring the heteromer or either homomer, (ii) oligomerization interfaces remain conserved, and (iii) no new function emerges. For a dimeric complex, the first two assumptions lead to 50% heteromer and 25% of each homomer at binding equilibrium right after the duplication ( Figure 2A ). We found that neutral fixation of mutations to this initial equilibrium favors the enrichment of the heteromeric fates ( Heteromeric or Mixed ) at the end of the simulation ( final equilibrium ) for >95% of the cases examined 96 . This is because due to their structural symmetry, the homomers accumulate destabilizing mutations (which are typically more abundant than stabilizing ones 97 , 98 ) faster than the heteromer. As this process continues, their binding affinities weaken, and larger fractions of the paralogs preferably heteromerize. Download figure Open in new tab Figure 2. Barriers against paralog heteromerization drive the enrichment of Homomeric paralogs . ( A ) Duplication of a homomeric protein’s gene introduces a statistical mixture of homo- and heteromers. Assuming identical synthesis rates of each paralog, identical binding affinities of each complex, and stochastic oligomerization, the initial equilibrium includes 25% of each homomer and 50% heteromer. This changes upon further divergence, as the accumulating mutations alter these factors. Typically, mutational biases under a neutral model lead to the enrichment of the heteromer. ( B ) Simulation setup from Cisneros et al. 96 , modified to include a barrier to paralog heteromerization. Each paralog is assigned a characteristic synthesis and decay rate and a folding free energy. The folded monomers are assigned a separation efficiency ( q ), which causes q % of them to only homomerize, while the remaining copies homo-/heteromerize by random encounters. The paralogs’ subsequent evolutionary divergence was simulated by introducing point mutations, with effects on folding free energies and binding affinities predicted by FoldX 99 , 100 . These mutations alter the concentration of folded monomers and also the affinities of their homo/hetero-dimeric complexes. ( C ) This simulation framework shown in panel B was applied to 37 high-quality homodimers with crystallographic structures. The percentages of heterodimers obtained at the final equilibrium (median of 50 replicates, 200 mutations fixed for each) for different separation efficiencies are plotted as a heatmap. ( D ) The distribution of %heteromers in the final equilibrium for different separation efficiencies from panel C are plotted as violin plots. The %heteromers in the initial equilibrium are highlighted by red dots. The boxes and whiskers indicate the 25 th -75 th percentiles and values that are 1.5 times the interquartile range. Our observations are at odds with this model, since homomeric fates are observed with a higher proportion (∼45%, Figure 1A ) than expected (<5%) from simulations, suggesting that one or many of the assumptions are not met in reality. These observations imply that underlying factors increase the likelihood of homomer formation. This could occur if certain barriers specifically hinder paralog heteromerization, causing homomers to become the dominant functionally active complexes in the initial equilibrium . To reflect such barriers, we modified the simulation framework to add constraints ( Figure 2B , Methods ). Each paralog was assigned a separation efficiency ( q ) representing that q % of the folded monomers only formed homomers. The remaining folded monomers could form both homo- and heteromers by random encounters according to their respective binding affinities (Δ G bind ). This separation efficiency thus acts as a barrier to heteromerization. As stronger separation efficiencies are applied, even if only on one paralog, heteromers are depleted in the initial equilibrium even if they have a stronger binding affinity than both homomers ( Figure S2 ). We applied this framework using various separation efficiencies ranging from 0 (no separation) to 1 (complete separation) to 37 homomeric structures ( Methods ) obtained from the Protein Data Bank (PDB) 70 . These structures were sampled from major ECOD architecture types 101 , and thereby represent independently evolved protein folds. In an iterative process, amino acid substitutions were introduced to the two paralogs that altered the Δ G fold of their monomeric states and the Δ G bind of their homo-/heteromeric complexes. Hence, the relative abundances of the mono- and oligomeric states changed upon each mutation. These mutations were randomly sampled from a pool of all possible single amino acid substitutions with effects on the folding free energies and the binding affinities estimated using FoldX 99 , 100 . Each simulation was replicated 50 times, with each replicate continuing until 200 substitutions were fixed. For simplicity, identical separation efficiencies were assigned to each paralog. Results show that as long as the barrier against paralog heteromerization was weak (0.0 ≤ q q > 0.4) was sufficient to favor homomer evolution. For example, the homodimeric FMN-binding protein (PDB code: 3ZOF 102 ) forms 86% heteromers for q = 0.0, which reduces to 50% for q = 0.4 and to 30% for q = 0.7. Higher separation efficiencies further decreased the proportion of heteromers at the start of the simulation and reduced their maximum attainable concentration. When q = 1, the two paralogs could not interact at all, and no heteromer was formed. However, for q < 1, the two paralogs could form a heteromer from the beginning of the simulation. As simulation continued and mutations accumulated, the proportion of heteromer gradually increased relative to that in the beginning ( Table S2 ), albeit these increases became smaller in magnitude with increasing values of q ( Figure 2D ). In our simulations, the two paralogs may accumulate mutations at their oligomeric interfaces, but the overall interface patch remains at the same position on the structure. As such, these simulated paralogous pairs, even after evolving independently in the presence of a barrier, selection for the maintenance of homomerization makes them retain the ability to heteromerize if the barrier was lifted ( Table S3, Figure S3 ). This is consistent with known paralogous pairs that localize to different subcellular compartments yet heteromerize in vitro 74 . Taken together, these results highlight that if oligomerization occurs by random encounters of monomers, then mutational biases combined with random genetic drift favors heteromerization of paralogs. Thus, the emergence of homomeric paralogs requires that barriers hinder the heteromeric interaction. With such barriers homomers can be favored, but what could these barriers be? Co-translational assembly is a major barrier promoting homomeric interactions We systematically quantified the contribution of different protein properties that can act as barriers to heteromer formation, and thus promote homomeric paralogs in human cells. These barriers may have multiple independent origins and could either be acquired de novo or be inherited from the pre-duplication ancestor ( Figure 3A ). Acquired barriers may be physical, such as the two paralogs being located in different subcellular compartments, or being expressed in different sets of tissues and/or cell types 56 . Alternatively, they could be temporal, as with a lack of gene co-expression 56 . Acquired barriers can also be structural, such as InDels at oligomeric interfaces, which are known to efficiently modulate protein-protein interactions 56 , 103 , 104 . Download figure Open in new tab Figure 3. The effect of different barriers on the oligomeric state divergence of human paralogs. ( A ) Different types of barriers promoting homomeric paralogs. ( B-G ) Violin plots comparing the percentages of Homomeric (orange), Heteromeric (green), and Mixed (yellow) paralogous pairs, per family per ancestral branch, in the presence (left) vs . in the absence (right) of different barriers. The numbers represent the total number of corresponding duplication events across the eukaryotic lineage. Each violin includes 22 data points (each corresponds to an ancestral branch); violin features follow Figure 2D . Differences between the distributions were assessed by two- sample t-tests. The barriers are depicted as follows: ( B ) different vs . common (co-occurring in at least one) subcellular compartments, ( C ) expression in different vs . common sets of tissues given by intersection/union ( IoU ) in a total of 253 RNAseq-based tissue-specific expression datasets ( IoU < 0.25 vs. IoU ≥ 0.25), ( D ) expression in different vs . common cell types ( IoU < 0.25 vs. IoU ≥ 0.25, for a total of 1056 RNAseq based cell-type-specific expression datasets), ( E ) not coexpressed vs . coexpressed paralogs as measured by the Spearman correlation ( R ) of their mRNA abundances across 253 RNAseq-based tissue-specific expression datasets ( R < 0.4 vs. R ≥ 0.4), ( F ) paralogs with vs . without interface InDels, ( G ) co-translational vs . post-translational assembly. ( H ) Barplots compare the strengths of different barriers as the deviation (Δ) of the average % Homomeric paralogs (per family per branch) observed in its absence vs . its presence. To quantify the effects of acquired barriers, we integrated multiple types of omics data together with the reconstructed histories of the duplicates ( Methods ). Protein subcellular localization as well as tissue and cell-type-specific gene expression data ( Data S3 ) were obtained from the Human Protein Atlas 105 . We first examined paralogous pairs that co-localize in the same common subcellular compartment vs . those that do not. For each category, the percentages of Homomeric, Heteromeric , and Mixed pairs per family were computed for each ancestral branch of the species tree. We found that Homomeric and Heteromeric fates show opposite trends. Homomer representation decreases from 54% among paralogous pairs that do not co-localize in the same subcellular compartment to 41% among pairs that do co-localize ( p = 0.002, Figure 3B ). By contrast, the Heterome r representation correspondingly increases from 22% to 37% ( p = 0.003, Figure 3B ). Similar trends were observed for paralogous pairs expressed in different vs . common sets of tissues ( Homomer representation decrease from 66% to 40%, p = 2e-05, Figure 3C ) and cell-types ( Homomer representation decrease from 62% to 38%, p = 1e-05, Figure 3D ). Also for temporal barriers we found that Homomer representation decreases from 55% among weakly co-expressed paralogous pairs to 41% among strongly co-expressed pairs ( p = 0.001, Figure 3E ). To examine the role of structural barriers, we used AlphaFold Multimer 106 , 107 to model quaternary structures of homo- and heteromeric paralogs ( Methods ). A total of 3760 homomeric and 955 heteromeric model structures ( Data S4 ) were examined to identify paralog-specific InDels at the oligomeric interfaces (both small InDels and domain gain/losses). We found that Homomer representation decreases from 76% among paralogs with interface InDels to 44% among those without ( p = 3e-13, Figure 3F ). In the same categories, the Heteromer representation correspondingly increases from 6% to 34% ( p = 2e-15, Figure 3F ). Importantly, alterations of gene expression or protein subcellular localization are not expected to be effective immediately after the duplication. Therefore, the pronounced increase in homomer representation in the presence of these barriers could be the consequence of the fate rather than the driver of the fate (oligomeric state divergence first versus barrier first scenarios). However, we reasoned that another barrier, co-translational assembly, would exist immediately after duplication. Co-translationally assembling homomers begin to interact as they emerge from their respective ribosomes 108 – 110 . For homomers, when both ribosomes translate the same mRNA ( cis -assembly), this mechanism creates a natural barrier to heteromerization. In this scenario, two paralogs are synthesized as two independent homomers even if they are co-expressed and co-localized in the same subcellular compartment. Moreover, we recently showed that co-translational assembly originates in structural and biophysical features of protein complexes and typically involves subunits that are geometrically intertwined and partner-stabilized 111 . Because these subunits tend to be unstable in isolation, we expect the paralogs’ post facto heteromerization by subunit exchange to be rare as well. Bertolini et al. 112 identified thousands of human proteins undergoing co-translational assembly and these could be accurately predicted from their structural characteristics 111 . Based on this integrated dataset, ∼17% of the human proteome was found to assemble co-translationally and ∼53% was annotated as assembling post-translationally. Projecting the reconstructed history of duplicates onto these data revealed that Homomer representation decreased from 67% among co-translational paralogs to 28% among post-translational paralogs ( p = 2e-11, Figure 3G ). In the same categories, the Heteromer representation increased from 13% to 49% ( p = 1e-10, Figure 3G ). This result provides conclusive evidence that co-translational self-assembly is a major determinant of paralogs’ homomeric fates. To compare the strength of barriers we used the difference (Δ) between the average % Homomeric fates in the presence vs . absence of a barrier ( Figure 3H ). We found that the temporal barrier (lack of gene co-expression) was the weakest, likely because human cells include ∼1000-fold variation of protein half-lives 113 – 115 and so protein products of weakly co-expressed genes can still interact if they happen to be long-lived. Physical barriers are of intermediate strength (tissue/cell-type-specific expression being stronger than subcellular localization). However, co-translational assembly appeared to be the strongest while interface InDels (structural barrier) was the second strongest. Taken together, these results provide conclusive evidence that homomeric paralogs dominate when certain barriers specifically hinder paralog heteromerization. In particular, co-translational assembly, originating from the structural features of protein complexes, appeared as the strongest barrier promoting homomeric paralogs in human cells and throughout eukaryotic evolution. Heteromerization constrains paralogs’ sequence, structural, and functional divergence To examine whether heteromerization of paralogs constrains their functional divergence, we compared homo- and heteromeric paralogs using three complementary measures of functional divergence ( Methods ). First, for paralogous enzymes we analyzed their Enzyme Commission (EC) numbers, which represents the biochemical reaction they catalyze. EC number differences thus represent altered substrate-specificity or reaction chemistry 116 . EC numbers of 3758 human enzymes obtained from Expasy 116 and KEGG 95 were mapped to 1622 paralogous pairs ( Data S3 ). We found that the average frequency of unique EC numbers per family ( i . e . one paralog catalyzes a reaction the other one doesn’t) increases with duplication age more rapidly for Homomeric pairs than for Heteromeric and Mixed (t-test p < 2E-05, Figure 4A ). Download figure Open in new tab Figure 4. Homomeric paralogs increasingly diverge in sequence, structure, and function with duplication age. ( A - C ) Ribbon plots represent the functional divergence of Homomeric (orange), Heteromeric (green) and Mixed (yellow) paralogous pairs. The lines and ribbons represent the mean and standard deviations. For clarity, each curve was smoothed using a sliding window approach (window size = 3). ( A ) The count of novel EC numbers per family. ( B ) The count of novel non-paralogous PPI partners per family. ( C ) The count of novel GO categories per family. ( D-F ) Same as panels A-C , for sequence and structural divergence of paralogous pairs. ( D ) The average frequency of substitutions, per 100 amino acids, per family. ( E ) The average frequency of InDel sites, per 100 amino acids, per family. ( F ) The average TM-score (of the structural superposition) between AlphaFold2-modeled structures of paralogous pairs, per family. ( G - I ) Genes/codons are assigned to evolve in adaptive or nearly-neutral regimes by comparing their nonsynonymous ( dN ) vs . synonymous substitution rates ( dS ). ( G ) Bar plots depicting the %paralogous genes in the adaptive regime. ( H ) Violin plots comparing the %codons per gene in the adaptive regime. Differences between the distributions were assessed by two-sample t-tests. ( I ) Same as panel- H , for the %codons per gene in the nearly-neutral regime. ( J - L ) Same as panels A-C , for the difference (Δ) of evolutionary rates ( dN / dS, dN , and dS are plotted along the Y-axis in panels D, E and F ). ( M - O ) Same as panels A-C , for functional divergence of co- (gray) and post-translational paralogs (blue). Next, we examined how paralogous pairs diverge to interact with unique unrelated partners. To that end, we mined 22184 heteromeric PPIs between co-complex subunits 72 , 73 ( Data S1 ), which included interaction data for 1480 paralogous pairs. The average frequency of unique interaction partners per family ( i . e . one paralog interacts with an unrelated partner the other one doesn’t) also increases with duplication age more rapidly for Homomeric pairs than Heteromeric and Mixed ( p < 8E-07, Figure 4B ). Finally, Gene Ontology (GO) data was analyzed to explore how paralogous pairs diverge to perform new functions. Here, GO terms related to ‘biological process’ and ‘molecular function’ 117 , 118 were obtained for 19070 human proteins and mapped onto 2245 paralogous pairs ( Data S3 ). The average count of novel GO terms per family ( i . e . one paralog is assigned a GO term that the other one isn’t) also increases with duplication age more rapidly for Homomeric pairs than Heteromeric and Mixed ( p < 3E-04, Figure 4C ). We asked whether the accelerated functional divergence of homomeric paralogs is coupled with their rapid sequence and structural divergence. Analyzing pairwise sequence alignments of paralogous pairs ( Methods ) we found that accumulation of substitutions and InDels increases with duplication age more rapidly for Homomeric pairs than for Heteromeric and Mixed ( p < 3E-06, and p < 2E-08, Figure 4D-E ). Furthermore, homomeric paralogs accumulate InDels and substitutions both at the interface and surface, whereas heteromeric paralogs predominantly keep their interfaces conserved ( Figure S4 ). This accelerated InDel accumulation is consistent with a faster divergence of protein structure and function in homomeric paralogous pairs 56 , 103 , 104 , 119 – 122 . Indeed, assessing structural divergence of paralogous pairs by the template-modeling score (TM-score 123 ) shows that Homomeric pairs diverged more rapidly than Heteromeric and Mixed pairs ( p < 0.01, Figure 4F ). Interestingly, Mixed pairs remained structurally more similar than Heteromeric ( p < 3E-08). Together, these results highlight that the rapid functional divergence of homomeric paralogs is associated with their accelerated sequence and structural divergence. Rapid accumulation of substitutions and acquisition of new functions are often driven by positive selection. Hence, we compared the nonsynonymous ( dN ) and synonymous substitution rates ( dS ) of paralogous genes seeking signatures of positive selection or neutral drift. The ratio of the two, called ω (or dN / dS ), provides a measure of the strength of selection exerted at the amino acid level 124 , 125 . In principle, ω is expected to be =1, 1 under neutral, purifying, or positive selection scenarios 126 – 128 . However, in practice, most proteins evolve under a mix of positive and purifying selection and are dominated by the latter, typically leading to ω ω 0 and ω ≈ ω 0 highlight evolution in the adaptive and nearly-neutral regimes. We used a dataset 129 annotating the evolution regime ( e . g ., adaptive and nearly-neutral) of genes and codons in the human genome. At the gene level, significantly more homomeric paralogs (∼10% versus ∼6% for heteromers) were observed to evolve in the adaptive regime ( Figure 4G ). Extending our analysis to the codon level, we found that Homomeric paralogs showed a higher percentage of codons in the adaptive regime when compared to the Heteromeric and Mixed categories ( p < 3e-25, Figure 4H ). Conversely, Heteromeric and Mixed paralogs harbored a higher percentage of codons in the nearly-neutral regime when compared to the Homomeric category ( p < 8e-23, Figure 4I ). These results highlight that homomeric and heteromeric paralogs are enriched in genes and codons evolving in the adaptive and nearly-neutral regimes respectively. We asked whether the functional divergence of Homomeric pairs is coupled to a pronounced sequence divergence. To that end, we compared the difference (Δ) of ω for homo- and heteromeric paralogs. Results show that Homomeric pairs typically display significantly higher differences ( i . e . more dissimilar ω ) than Heteromeric and Mixed pairs of the same duplication age ( p < 2E-10, Figure 4J ). These trends were recapitulated for dN ( p 0.2, Figure 4L ), highlighting that they correspond to changes in the protein sequence, as previously reflected in Figure 4D-E . Together, these results highlight that homomeric paralogs tend to evolve more asymmetrically (functional divergence plus dissimilar evolutionary rates) than heteromeric paralogs. Does the asymmetric evolution of Homomeric paralogs kick in immediately after the duplication? As depicted in Figure 4A-C and J-L , it appears that pronounced differences in EC numbers, interaction partners, molecular functions, and evolutionary rates are primarily found between homo- and heteromeric paralogs that duplicated before the Boreoeutheria node (node A5, placental mammals, originated 94(±5) million years ago in mid-Cretaceous 66 , 130 ). Hence, a ‘lag phase’ of ∼100 million years appears to exist between the duplication event and the onset of asymmetric evolution. Interestingly, as depicted in Figure 4M-O , this lag is relatively shorter in co-translational homomeric paralogs (inherited barrier, effective immediately) as compared to post-translational (acquired barriers, require evolutionary change accumulation). Thus, the onset of asymmetric evolution appears to follow the apparition of the barrier to heteromerization. These results show that the asymmetric evolution of paralogous pairs is tied to their structural and biophysical independence. Homomeric paralogs diverge more rapidly over time in enzymatic reactions, interaction partners, and molecular functions, while heteromeric paralogs remain more functionally linked. DISCUSSION We have shown that gene duplication of homomeric proteins often results in independent homomers if heteromerization is prevented by barriers. Co-translational self-assembly of each paralog appeared as an important such barrier. Interestingly, this barrier likely operates immediately after duplication because the structural properties that drive co-translational assembly are inherited from the ancestor. By contrast, other barriers such as spatiotemporal separation or structural incompatibility likely appear with a delay after the duplication event. In the absence of such barriers, the heteromer is favored. As they accumulate mutations, both paralogs could become ‘entrenched’ into a heteromeric complex, which would slow down their functional divergence 61 , 131 – 133 . This phenomenon, known as ‘constructive neutral evolution’ 134 is likely widespread in eukaryotes 135 , 136 and could have contributed to homomer-dominant prokaryotic proteomes transforming into heteromer-dominant eukaryotic proteomes 55 , 56 . Co-translational self-assembly appeared as the most significant barrier promoting homomeric fates ( Figure 3G,H ). In principle, co-translational self-assembly can occur either in cis or in trans ; that is, from two ribosomes translating the same or different mRNAs respectively. While a handful of experimentally characterized homomers were shown to assemble in cis 112 , 137 , 138 , which mode of assembly dominates in the cell, cis or trans , remained elusive. Our results are consistent with the dominance of cis -assembly in eukaryotic genomes, and this trend was also reflected in co-translational homomeric proteins being depleted in mutations associated with autosomal dominant disorders 137 . Many paralogs are known to evolve asymmetrically ( i . e . diverge in structure, function, and evolutionary rates), but what fraction of paralogs do so and why, is debated 139 – 144 . We showed that asymmetric evolution is coupled with paralogs’ structural and biophysical independence. These results are complementary to observations made for yeast whole-genome duplicates 26 . In summary, we unveiled a fundamental connection between co-translational assembly and the evolutionary divergence of protein oligomeric state and function. The resulting framework highlights the power of combining phylogenomics with multi-OMICs data and further helps resolve long-standing questions such as why paralogs appear to reach specific oligomeric fates, why only some paralogs evolve asymmetrically, and why temporal lags exist between whole genome duplications and subsequent adaptive radiation. Author contributions S.M., A.F.C., C.R.L., and E.D.L. conceived and conceptualized the research. S.M. performed the phylogenetic, OMICs, and evolutionary rate analyses shown in Figures 1 , 3 , and 4 . A.F.C. performed the biophysical simulations shown in Figure 2 , structural analysis shown in Figure 3G and 4F , and conceptualized the barrier strength analysis shown in Figure 3H . A.F.C. also generated the AlphaFold2 models of heteromeric paralogs. S.M. and A.F.C. prepared the figures and wrote the first draft. E.D.L. and C.R.L. acquired funding and supervised the study. All authors discussed and commented on the results and edited versions of the paper. Competing interests The authors declare no competing interests. Data and Materials availability All data used or generated in this study are provided as Supplementary Materials. AlphaFold2 models of paralogous protein complexes are available in Zenodo: https://doi.org/10.5281/zenodo.14158203 . Materials and Methods Taxon sampling and species tree construction We sampled 123 organisms with fully sequenced and annotated genomes ( Table S1 ) following our previously published methodology 145 . Briefly, for all major phylogenetic clades of the domain Eukaryota, one or more representative species were selected, which included a total of 106 taxa. Aiming to uniformly sample the Eukaryota, we ensured that any pair of species in our dataset has a minimum divergence time of 65 million years, with divergence time data obtained from TimeTree v5 66 . In addition, we recruited 17 outgroup species, comprising 11 and 6 species representing various Proteobacterial and Asgard archaea classes respectively. The TimeTree database v5 comprises the phylogenetic relationships among 137306 species and their divergence times estimated by the molecular clock analyses calibrated by fossil records 66 . Our species set was submitted to the TimeTree database to obtain the eukaryotic Tree of Life which served as the species tree. The tree leaves represent the extant species and the internal nodes represent their now-extinct ancestors. The tree topology was manually adjusted to include the Proteobacterial and the Asgard archaea clades as the outgroups since their endosymbiosis gave rise to the first eukaryote 146 . The branch lengths represent the evolutionary divergence times (million years) as documented in TimeTree. Due to the near-absence of fossil records, for ancestral branches older than Metazoa , the divergence time estimates are based solely on molecular dating and are therefore putative. Genome-wide comparative phylogenomic analysis Identifying protein families Annotated proteome sequences of the 123 organisms were obtained from UniProt 147 , including each gene’s canonical protein sequence. Gene families were identified using the OrthoFinder package v2.5.5 69 . OrthoFinder analyses include two steps. In the first step, protein sequences were compared in all- vs .-all BLASTp searches 67 . The resulting BLASTp hits were filtered by removing all query-subject-pairs for which the alignment coverage was <30% of either the total query or subject sequence length. This step is expected to significantly improve phylogenetic trees and orthology inference in the subsequent steps 148 , 149 . In the second step, based on these modified BLASTp tables the remaining 2162331 sequences were clustered (inflation parameter of 1.5) into 60259 protein families which were considered proxies of protein families. Filtering protein families that have descended from homomeric ancestors A total of 5330 families, out of the initial 60259, included at least two human paralogs. To identify which of these families are descendants of ancestral homomeric proteins, we examined the molecular interaction status of the human paralogs. To that end, a curated dataset of 4940 homomeric and 142226 heteromeric interactions (see “ Compiling molecular interaction data ”) was mapped to these paralogous pairs. Each paralogous pair was classified into one of the three categories, Homomeric (both paralogs are homomeric and they do not heteromerize), Heteromeric (the two paralogs form a heteromer, and no homomeric interaction was observed), or Mixed (both paralogs are homomeric and they also form a heteromer). Paralogous pairs of these three categories are parsimoniously expected to have emerged from ancestral homomeric proteins 54 – 57 . We identified 1998 protein families, each including at least one such paralogous pair ( Data S2 ). These 1998 families include 7377 human proteins. Multiple sequence alignment and trimming Protein sequences of each family were aligned by the 3D-Coffee alignment tool, implemented in T-Coffee package 150 . 3D-Coffee generates structure-guided sequence alignments. Briefly, 3D structures of multiple template proteins are aligned to create a multiple structural alignment, which is then used as a guide to align the remaining sequences. We collected proteome-scale AlphaFold2-modeled 151 monomeric structures of 10 organisms ( Homo sapiens, Mus musculus, Danio rerio, Drosophila melanogaster, Caenorhabditis elegans, Arabidopsis thaliana, Saccharomyces cerevisiae, Plasmodium falciparum, Trypanosoma brucei , and Escherichia coli ). These structures were used as the templates. These multiple alignments were subsequently trimmed using the trimAL tool 152 (parameter settings: -gt 0.6 -cons 50) to remove gap-majority columns. Phylogenetic tree inference For each of the 1998 protein families, command-line Mega v11.0.8 was used to infer Maximum-Likelihood unrooted protein trees (model = LG+F, Rates among Sites = G+I, number of discrete gamma categories = 5, ML Heuristic Method = SPR level 5, Initial Tree = NJ/BioNJ, Test of Phylogeny = 100 Bootstrap Replicates). These trees are available in newick format in Data S2 . Compiling molecular interaction data Complexes with experimental/predicted structures We obtained structures of 837 human homomeric complexes resolved by X-ray crystallography or Electron Microscopy (EM) from the 3DComplex database 153 , 154 , based on the following criteria. (i) We selected homomers whose QSBio error rate was below 15% 153 . (ii) We imposed a minimum sequence coverage of ≥ 70% for the respective UniProt sequence 147 . (iii) Structures corresponding to artificial fusion proteins (N = 33) or proteins encoded by transposon elements (N = 12) were discarded. Additionally, we used the results from our recently developed AlphaFold2-based 106 framework that predicts homomeric structures across full proteomes 71 . These predictions enabled us to map a high-quality dataset of homomeric complex structures for 3584 human proteins. For heteromers, we collected a set of 2606 heteromeric complex structures from Protein Data Bank 70 , from which subunit composition and inter-subunit interaction data were extracted. Curated macromolecular complexes Subunit protein composition data for 1265 and 2303 experimentally annotated human heteromeric complexes were obtained from Complex Portal 72 and hu.MAP v2.0 73 databases. Protein-protein interaction data Human protein-protein interaction (PPI) data, including both homo- and heteromeric interactions were obtained from seven databases. This initial dataset included 73834, 142706, 64923, 181555, 51503, 114654, and 173350 interacting pairs obtained from BioGrid 74 , IntAct 75 , HuRI 76 , Mentha 77 , MINT 78 , HiNT 79 , and HIPPIE 80 . In total, these databases included 390709 unique PPI interactions. Since PPI data tend to be relatively noisy, we subsequently filtered these interactions as described previously in ref. 55 . Briefly, in the first step, all predicted interactions and text-mining-based interactions were removed. In the second step, to minimize false positives, we demanded that the interaction between two proteins be observed using both proteins as bait and as prey. These filtering steps provided a high-quality dataset of 192543 interactions. A comprehensive human interactome We aimed to unify these different datasets together to obtain a comprehensive set of human homomeric proteins and heteromeric protein pairs. A homomeric interaction was included in the final dataset (i) if the interaction was observed in at least one X-ray crystallographic or EM or AlphaFold2-predicted structure, or (ii) if the interaction was observed in at least three independent PPI experiments. These filtering steps provided a dataset of 4940 human homomeric proteins ( Data S1 ). A heteromeric pair was included in the final dataset, (i) if their pairwise interaction was observed in at least one X-ray crystallographic or EM or AlphaFold2-predicted structure, or (ii) if both are members of the same curated complex, and their pairwise interaction was observed in at least one PPI experiment, or (iii) if their pairwise interaction was observed in at least three independent PPI experiments. These filtering steps provided a final dataset of 142226 human heteromeric protein pairs ( Data S1 ). Annotating gene duplication events We annotated gene duplication events on the species tree by two complementary approaches. Gene tree-species tree reconciliation approach We used Notung v2.9 155 to reconcile the protein trees of the 1998 protein families with the species treeusing the duplication-transfer-loss model (-- rearrange --threshold 70.0 --costdup 1.5 --costtrans 1.0 --costloss 1.0 --prune --treeoutput nhx -- events). Given a protein tree and a species tree, Notung infers gene duplications, losses, and transfers along a species tree. For each inferred duplication event, the bootstrap support of the respective branch in the protein tree was considered as the statistical confidence of the inference. Maximum-Likelihood-based inference of ancestral gene copy numbers We annotated gene duplication events using a complementary approach which is based on inferring the gene copy numbers for the ancestral nodes of the species tree. We used PastML 81 , which infers ancestral characters on the internal nodes of a rooted phylogenetic tree (species tree) with annotated tips (copy- number annotation) using maximum-likelihood methods (ML method = Marginal Posterior Probabilities Approximation or MPPA, character evolution model = Estimate-from-Tips or EFT). A posterior probability score is assigned to each copy-number inference, which was considered as the statistical confidence for inferring duplication events. For the 1998 families, we inferred a total of 5381 duplication events, which were subsequently filtered based on the following criteria. First, we only considered those duplications for which Notung-based and PastML-based inferences converged, i . e ., they both traced the duplication event to the same ancestral branch. Additionally, only high-confidence duplication inferences (those inferred with ≥ 70 bootstrap support or with ≥ 0.7 posterior probability) were kept. These filtering steps provided a final set of 4386 high-confidence duplications ( Data S2 ). Assigning the oligomeric fates of ancestral duplications Given the oligomeric states of the extant human paralogous genes, the fates of ancestral duplications were assigned by maximum parsimony using PastML 81 . The results are available in Data S2 . Simulating Oligomeric State Evolution Selection of protein structures A total of 37 structures obtained from Protein Data Bank (PDB) 70 was analyzed ( Table S2, S3 ); this dataset was compiled in our previous work 96 . Briefly, we mined PDB for homodimeric structures resolved by X-ray crystallography composed of two identical subunits 60-450 amino acids in length. Structures were sampled from major ECOD architecture types 101 , each of which represents an independently evolved protein fold. To remove redundancy, we clustered the resulting 104 structures using CD-Hit version 4.8.1 156 , 157 at 40% sequence identity. All structures were identified in different clusters. Finally, we selected 37 structures that were annotated in the highest quality tier in the QSBio database 153 (error probability < 2%). Estimation of mutational effects Effects of point mutations on each homomeric structure were estimated as in our previous work 96 . Briefly, the biological assemblies of the 37 structures were energy-minimized using the FoldX Repair function 99 ten times to ensure convergence to an energy minimum 158 . Then, in silico mutagenesis was performed using the FoldX BuildModel and AnalyseComplex functions 99 with the MutateX workflow 100 . The effects of mutations on the folding free energy and the binding affinity of heterodimers were calculated by simulating mutations on only one of the protein copies. Effects on the binding affinity of the homodimer were obtained by simulating mutations on the two copies. Framework for simulations Simulations were initialized with a gene encoding a homomeric protein with typical values of folding free energy 159 , 160 and binding affinity 161 , 162 . Folding free energy is used to calculate the fraction of properly folded subunits following equation 1 163 : where w is the fraction of folded proteins and Δ G fold is the free energy of the native fold. Properly folded subunits were then assigned a separation efficiency ( q ), representing the fraction of protein copies that can only form homomers. The rest of the protein copies were allowed to mix, thus forming both homo- and heterodimers based on their binding affinities. Equilibrium concentrations of monomers, homodimers, and heterodimers were then calculated using the same system of equations as in our previous work 96 . Equations 2 and 3 allow calculating the equilibrium concentrations of monomers and homomers for the fraction of protein copies that can only homomerize: where c A is the concentration of monomers, c AA is the concentration of homodimers, d A and d AA are the decay rates of monomers and homodimers, s A is the rate of synthesized copies of protein A that can only homomerize, and k AA is the association constant for the homodimer. The same equations are used for the equilibrium concentrations for monomers and homodimers of protein B. Similarly, equations 4 - 8 were used to calculate the equilibrium concentrations for the fraction of paralogs that were allowed to mix: Decay rates were set to 1.3 h −1 based on estimates for the half-life of proteins and cell division rates in yeast. The fitness function used 60 h −1 as the optimum, allowing us to simulate the case in which it is overshot by the total activity after duplication (76.4 h −1 ). Simulations continued by sampling from the distributions of mutational effects for each of the selected structures until 200 mutations were fixed, with 50 replicates of each simulation. Each sampled mutation affects the folding free energy and the binding energy of the homo- and heterodimers, which modifies their concentrations at equilibrium. Annotating barriers to paralog heteromerization Co-/Post-translational assembly data We used the results from our recently developed framework that integrated structure-based predictions with experimental characterization to gain an overview of co-translational assembly across the Homo sapiens, Saccharomyces cerevisiae , and Escherichia coli proteome 111 . In all three organisms, we found that co-translationally assembling complex subunits represent ∼20% of the proteome whereas ∼50% was annotated as post, and the remaining ∼30% could not be annotated due to ambiguous or absent data. These genome-scale annotations were mapped to the 1998 protein families. A protein family was annotated as co- or post-translational if all the mapped family members were either co- or post-translational. This analysis enabled us to annotate 591 co-translational and 792 post-translational families ( Data S3 ). Protein subcellular localization data Subcellular localization data for 13106 human proteins were obtained from the Human Protein Atlas 105 and were subsequently filtered to remove uncertain annotations. The final dataset included high-confidence subcellular localization data for 12018 human proteins. Each protein was identified by Ensembl IDs and was mapped onto the human reference proteome according to UniProt mapping ( Data S3 ). Gene expression data We obtained the RNAseq-based presence/absence data of 19619 human transcripts in 257 tissues and 1056 cell types from the Human Protein Atlas 105 . We also collected the corresponding processed mRNA abundance data (Transcripts per Million, or TPM values). The mRNAs were identified by Ensembl IDs and were mapped onto the human reference proteome according to UniProt mapping. In cases where multiple transcripts matched one UniProt ID, only the one with the highest mean abundance was kept ( Data S3 ). Interface evolution of paralogous pairs A three-step approach was undertaken to examine the interface evolution of paralogous pairs. (i) AlphaFold2 models for homomers and heteromers We compiled a dataset of AlphaFold2-modeled structures of homo- and heteromeric human paralogs. A total of 3760 homomeric models were re- used from our previous study 71 . Additionally, we generated models of 955 heteromeric pairs using AlphaFold Multimer v2.3 106 , 107 , using a custom ColabFold implementation 164 . First, for each of the 7377 homomer-derived human paralogous genes, we used MMseqs2 165 to identify their orthologs in 3788 UniProt-annotated eukaryotic proteomes (1 st March, 2023). Those with ≥ 80% query sequence coverage (and ≥ 40% target sequence coverage) and 30% ≤ identity ≤ 80% were considered as high- quality hits. The sequence identity upper bound is expected to increase the number of effective sequences from which AlphaFold2 would extract the evolutionary information. From each proteome, we only considered the top hit. The hits for each paralogous pair were then compared, and only those with matching taxon were kept. The resulting hits were further filtered so that no two hits had ≥ 85% sequence identity with each other. The final sets of hits were considered high-quality orthologs. These sequences were then aligned with the Muscle v5.1 using the super5 algorithm 166 . These alignments were then fed to AlphaFold2, using the monomeric structures from AlphaFold database 151 as structural templates. Five heteromeric structures were generated for each of the 2077 paralog pairs, for a total of 10385 models. These models were generated for full-length protein sequences. As a result, these models often contain long flexible, disordered regions masking an otherwise structurally conserved core. These disordered regions were removed to extract these conserved core regions for each complex, using a computational pipeline adapted from Schweke et al. 71 . This pipeline also computes the probability score for each inferred complex being a physiological dimer. For heteromer, the model with the highest dimer probability was selected. Finally, a total of 878 models with ≥ 0.5 dimer probabilities and core structures of at least 60 residues for both subunits were considered for further analysis, which included 181 Heteromeric and 697 Mixed pairs ( Data S4 ). (ii) Benchmarking the AlphaFold2 models We used two benchmark tests to assess the accuracy of our heteromer models. First, on May 8th, 2023 we collected experimental structures of 90 heterodimers submitted to PDB after September 30, 2021, the AlphaFold Multimer v2.3 training date. MMseqs2 165 was then used to identify any potential sequence homologs (≥30% identity, ≥60% query coverage, ≥60 aa sequence length) of these structures that were present in the AlphaFold2 training set. We identified a total of 51 benchmark heterodimeric structures with no such potential homologs. These structures were not necessarily heteromeric paralogs. For each benchmark heterodimer, a model structure with dimer probability ≥0.5 was generated following the pipeline described above, and the two structures were subsequently aligned using the MM-align structural alignment tool (version 20191021) 167 . The reliability of modeled structures was assessed by their RMSD scores of alignment and DockQ scores 168 . Out of the initial 51, a total of 45 (90%) models accurately recapitulated the heteromeric interaction interface (DockQ ≥ 0.23), with 14 (28%) being high-quality predictions (DockQ ≥ 0.8) ( Figure S5 ). For a second benchmark, we compared our models to those produced by the AlphaFold3 server 169 . A total of 40 heteromeric paralogous pair models were selected for this analysis, their dimer probabilities spanning the entire range of zero to one. For each, an AlphaFold3 model was generated and then were pairwise compared as described above. Our AlphaFold2 models with dimer probability ≥ 0.5 had low RMSD values and high DockQ scores ( Figure S5 ) when compared to their AlphaFold3 counterparts. Thus, our filtered models are comparable to those obtained with AlphaFold3. These benchmark analyses confirm that our workflow generates and identifies reliable models for heteromeric protein-protein interactions. (iii) Interface analysis. Dimeric interfaces of our high-quality models were annotated following Levy 170 . Briefly, FreeSASA v2.0.3 171 was used to compute the solvent-accessible surface area ( ASA ) of each amino acid residue in the context of a Gly-X-Gly peptide, in the monomeric ( ASA m ) and also in the oligomeric state ( ASA o ). Indeed, ASA o < ASA m highlights that the residue is buried at the interface. To examine whether a particular interface residue site is conserved in both paralogs or not, we aligned their sequences using NEEDLE v6.6 172 . Comparing the interface residue annotations with the sequence conservation allowed us to measure the variations of the interface. An interface residue site was annotated as ‘conserved’ if it was found at the interface of both paralogs or as ‘novel’ if found at the interface of only one paralog. We used the sequence alignment to identify insertions and deletions (InDels) as well Sequence divergence of paralogous pairs Amino acid sequences of human paralogous pairs were aligned using Clustal Omega 173 and Muscle 166 . These pairwise alignments were used to detect the aligned positions, substitutions, and InDels. Structural divergence of paralogous pairs The core structures of high-quality AF2-derived models were trimmed to their core structures using adapted scripts from Schweke et al. 71 . The core structures of each pair of paralogs (3703 Homomeric , 697 Mixed , and 181 Heteromeric pairs) were then aligned using the MM-align structural alignment tool (version 20191021) 167 to calculate the template-modeling score (TM-score), normalized by the smaller of both paralogs, which is a metric of structural similarity. The TM-score ranges from 0 to 1 and values closer to 1 indicate higher structural similarity. Functional divergence of paralogous pairs Enzyme Commission number and reaction data Enzyme Commission numbers of 3758 annotated human enzymes were obtained from Expasy 116 and KEGG databases 95 . These enzymes mapped to 1622 paralogous pairs (1317 proteins, Data S3 ). An enzyme may have one or more EC numbers assigned to it and the latter includes promiscuous enzyme activity or bifunctional enzymes. For simplicity, we selected only one EC number per enzyme: the one that is the closest to its paralogous enzymes. Gene Ontology data We obtained Gene Ontology (GO) data for 19070 human proteins from GOA database 174 of EMBL-EBI, which included data for 2245 paralogous pairs. Since subcellular compartmentalization acts as a barrier against paralog heteromerization, to avoid circular arguments, GO terms related to subcellular locations were excluded, and those related to ‘biological process’ and ‘molecular function’ were considered in the analysis. Evolutionary rate data We obtained the rates of the rate of substitutions at synonymous ( dS ) and non-synonymous sites ( dN ) for 12348 human protein-coding genes from a previous systematic analysis of phylogenetic and population data 129 ( Data S3 ). In this study, the authors compiled comprehensive sequence datasets at phylogenetic scale. The various evolutionary rate parameters were measured by employing a Bayesian Mutation-Selection Framework 175 . The genes were identified by Ensembl IDs and were mapped onto the human reference proteome according to UniProt mapping. Acknowledgments We thank Yitzhak Pilpel, Sudip Kundu, Soham Dibyachintan, and Carla Bautista Rodríguez for helpful discussions. E.D.L. acknowledges support from the European Research Council under the European Union’s Horizon 2020 research and innovation program (grant agreement No. 819318), the Human Frontier Science Program (ref. no: RGP0016/2022), the Israel Science Foundation (grant No. 1452/18), and the Abisch-Frenkel Foundation. C.R.L. acknowledges support from Canadian Institutes of Health Research (CIHR) Foundation grant No. 387697, NSERC discovery grant, Canada Research Chair in Cellular Systems and Synthetic Biology. A.F.C. acknowledges support from the Fonds de recherche du Québec - Nature et technologies (dossier 290237), the Ministère de l’Enseignement Supérieur du Québec, the Agencia mexicana para la cooperación y el desarrollo internacional, a Mitacs Globalink Research Award (award number IT28316), and a PROTEO Graduate Scholarship. Footnotes ↵ † Co-first authors. Emails and ORCIDs A.F.C.: angel-fernando.cisneros-caballero.1{at}ulaval.ca https://doi.org/10.5281/zenodo.14158203 REFERENCES 1. ↵ Ohno , S. Evolution by Gene Duplication . ( Springer Science & Business Media , 2013 ). 2. Hakes , L. , Pinney , J. W. , Lovell , S. C. , Oliver , S. G. & Robertson , D. L. All duplicates are not equal: the difference between small-scale and genome duplication . Genome Biol 8 , R209 ( 2007 ). OpenUrl CrossRef PubMed 3. Lynch , M. , O’Hely , M. , Walsh , B. & Force , A. The probability of preservation of a newly arisen gene duplicate . Genetics 159 , 1789 – 1804 ( 2001 ). OpenUrl Abstract / FREE Full Text 4. ↵ Lynch , M. & Force , A. The probability of duplicate gene preservation by subfunctionalization . Genetics 154 , 459 – 473 ( 2000 ). OpenUrl Abstract / FREE Full Text 5. ↵ Maere , S. et al. Modeling gene and genome duplications in eukaryotes . Proc Natl Acad Sci U S A 102 , 5454 – 5459 ( 2005 ). OpenUrl Abstract / FREE Full Text 6. ↵ Lynch , M. & Conery , J. S. The evolutionary fate and consequences of duplicate genes . Science 290 , 1151 – 1155 ( 2000 ). OpenUrl Abstract / FREE Full Text 7. Takahata , N. & Maruyama , T. Polymorphism and loss of duplicate gene expression: a theoretical study with application of tetraploid fish . Proc Natl Acad Sci U S A 76 , 4521 – 4525 ( 1979 ). OpenUrl Abstract / FREE Full Text 8. Watterson , G. A. On the time for gene silencing at duplicate Loci . Genetics 105 , 745 – 766 ( 1983 ). OpenUrl Abstract / FREE Full Text 9. ↵ Bailey , G. S. , Poulter , R. T. & Stockwell , P. A. Gene duplication in tetraploid fish: model for gene silencing at unlinked duplicated loci . Proc Natl Acad Sci U S A 75 , 5575 – 5579 ( 1978 ). OpenUrl Abstract / FREE Full Text 10. ↵ Force , A. et al. Preservation of duplicate genes by complementary, degenerative mutations . Genetics 151 , 1531 – 1545 ( 1999 ). OpenUrl Abstract / FREE Full Text 11. He , X. & Zhang , J. Rapid subfunctionalization accompanied by prolonged and substantial neofunctionalization in duplicate gene evolution . Genetics 169 , 1157 – 1164 ( 2005 ). OpenUrl Abstract / FREE Full Text 12. Baker , C. R. , Hanson-Smith , V. & Johnson , A. D. Following gene duplication, paralog interference constrains transcriptional circuit evolution . Science 342 , 104 – 108 ( 2013 ). OpenUrl Abstract / FREE Full Text 13. Deng , C. , Cheng , C.-H. C. , Ye , H. , He , X. & Chen , L. Evolution of an antifreeze protein by neofunctionalization under escape from adaptive conflict . Proc. Natl. Acad. Sci. U. S. A . 107 , 21593 – 21598 ( 2010 ). OpenUrl Abstract / FREE Full Text 14. Bridgham , J. T. , Brown , J. E. , Rodríguez-Marí , A. , Catchen , J. M. & Thornton , J. W. Evolution of a new function by degenerative mutation in cephalochordate steroid receptors . PLoS Genet . 4 , e1000191 ( 2008 ). OpenUrl CrossRef PubMed 15. Kaltenegger , E. , Eich , E. & Ober , D. Evolution of homospermidine synthase in the convolvulaceae: a story of gene duplication, gene loss, and periods of various selection pressures . Plant Cell 25 , 1213 – 1227 ( 2013 ). OpenUrl Abstract / FREE Full Text 16. Prakashrao , A. S. et al. The long road of functional recruitment-The evolution of a gene duplicate to pyrrolizidine alkaloid biosynthesis in the morning glories (Convolvulaceae) . Plant Direct 6 , e420 ( 2022 ). OpenUrl CrossRef 17. ↵ Nowak , M. A. , Boerlijst , M. C. , Cooke , J. & Smith , J. M. Evolution of genetic redundancy . Nature 388 , 167 – 171 ( 1997 ). OpenUrl CrossRef PubMed Web of Science 18. Wagner , A. Asymmetric functional divergence of duplicate genes in yeast . Mol Biol Evol 19 , 1760 – 1768 ( 2002 ). OpenUrl CrossRef PubMed Web of Science 19. Wagner , A. The role of population size, pleiotropy and fitness effects of mutations in the evolution of overlapping gene functions . Genetics 154 , 1389 – 1401 ( 2000 ). OpenUrl Abstract / FREE Full Text 20. ↵ Walsh , J. B. How often do duplicated genes evolve new functions? Genetics 139 , 421 – 428 ( 1995 ). OpenUrl Abstract / FREE Full Text 21. ↵ Bourret , J. , Borvető , F. & Bravo , I. G. Subfunctionalisation of paralogous genes and evolution of differential codon usage preferences: The showcase of polypyrimidine tract binding proteins . J Evol Biol 36 , 1375 – 1392 ( 2023 ). OpenUrl CrossRef PubMed 22. Qiu , Y. , Van Tay , Y. , Ruan , Y. & Adams , K. L. Divergence of duplicated genes by repeated partitioning of splice forms and subcellular localization . New Phytol 225 , 1011 – 1022 ( 2020 ). OpenUrl CrossRef PubMed 23. González , J. et al. Diversification of Transcriptional Regulation Determines Subfunctionalization of Paralogous Branched Chain Aminotransferases in the Yeast . Genetics 207 , 975 – 991 ( 2017 ). OpenUrl Abstract / FREE Full Text 24. Turetzek , N. , Pechmann , M. , Schomburg , C. , Schneider , J. & Prpic , N.-M. Neofunctionalization of a Duplicate dachshund Gene Underlies the Evolution of a Novel Leg Segment in Arachnids . Mol Biol Evol 33 , 109 – 121 ( 2016 ). OpenUrl CrossRef PubMed 25. DeLuna , A. et al. Exposing the fitness contribution of duplicated genes . Nat Genet 40 , 676 – 681 ( 2008 ). OpenUrl CrossRef PubMed Web of Science 26. ↵ Kuzmin , E. et al. Exploring whole-genome duplicate gene retention with complex genetic interaction analysis . Science 368 , ( 2020 ). 27. ↵ Ezoe , A. , Shirai , K. & Hanada , K. Degree of Functional Divergence in Duplicates Is Associated with Distinct Roles in Plant Evolution . Mol Biol Evol 38 , 1447 – 1459 ( 2021 ). OpenUrl CrossRef PubMed 28. Marques , A. C. , Vinckenbosch , N. , Brawand , D. & Kaessmann , H. Functional diversification of duplicate genes through subcellular adaptation of encoded proteins . Genome Biol 9 , R54 ( 2008 ). OpenUrl CrossRef PubMed 29. Soria , P. S. , McGary , K. L. & Rokas , A. Functional divergence for every paralog . Mol Biol Evol 31 , 984 – 992 ( 2014 ). OpenUrl CrossRef PubMed Web of Science 30. De Smet , R. , Sabaghian , E. , Li , Z. , Saeys , Y. & Van de Peer , Y. Coordinated Functional Divergence of Genes after Genome Duplication in. Plant Cell 29 , 2786 – 2800 ( 2017 ). OpenUrl Abstract / FREE Full Text 31. ↵ Blanc , G. & Wolfe , K. H. Functional divergence of duplicated genes formed by polyploidy during Arabidopsis evolution . Plant Cell 16 , 1679 – 1691 ( 2004 ). OpenUrl Abstract / FREE Full Text 32. ↵ Krakauer , D. C. & Nowak , M. A. Evolutionary preservation of redundant duplicated genes . Semin Cell Dev Biol 10 , 555 – 559 ( 1999 ). OpenUrl CrossRef PubMed Web of Science 33. ↵ Thomas , J. H. Thinking about genetic redundancy . Trends Genet 9 , 395 – 399 ( 1993 ). OpenUrl CrossRef PubMed Web of Science 34. ↵ Diss , G. , Ascencio , D. , DeLuna , A. & Landry , C. R. Molecular mechanisms of paralogous compensation and the robustness of cellular networks . J Exp Zool B Mol Dev Evol 322 , 488 – 499 ( 2014 ). OpenUrl CrossRef PubMed 35. ↵ Kastner , P. , Mark , M. & Chambon , P. Nonsteroid nuclear receptors: what are genetic studies telling us about their role in real life? Cell 83 , 859 – 869 ( 1995 ). OpenUrl CrossRef PubMed Web of Science 36. Rudnicki , M. A. , Braun , T. , Hinuma , S. & Jaenisch , R. Inactivation of MyoD in mice leads to up-regulation of the myogenic HLH gene Myf-5 and results in apparently normal muscle development . Cell 71 , 383 – 390 ( 1992 ). OpenUrl CrossRef PubMed Web of Science 37. ↵ Saga , Y. , Yagi , T. , Ikawa , Y. , Sakakura , T. & Aizawa , S. Mice develop normally without tenascin . Genes Dev 6 , 1821 – 1831 ( 1992 ). OpenUrl Abstract / FREE Full Text 38. Wang , Y. , Schnegelsberg , P. N. , Dausman , J. & Jaenisch , R. Functional redundancy of the muscle-specific transcription factors Myf5 and myogenin . Nature 379 , 823 – 825 ( 1996 ). OpenUrl CrossRef PubMed 39. ↵ Joyner , A. L. , Herrup , K. , Auerbach , B. A. , Davis , C. A. & Rossant , J. Subtle cerebellar phenotype in mice homozygous for a targeted deletion of the En-2 homeobox . Science 251 , 1239 – 1243 ( 1991 ). OpenUrl Abstract / FREE Full Text 40. ↵ Taniguchi , T. Cytokine signaling through nonreceptor protein tyrosine kinases . Science 268 , 251 – 255 ( 1995 ). OpenUrl Abstract / FREE Full Text 41. ↵ Steindler , D. A. et al. Tenascin knockout mice: barrels, boundary molecules, and glial scars . J Neurosci 15 , 1971 – 1983 ( 1995 ). OpenUrl Abstract / FREE Full Text 42. ↵ Reed , S. I. G1-specific cyclins: in search of an S-phase-promoting factor . Trends Genet 7 , 95 – 99 ( 1991 ). OpenUrl CrossRef PubMed Web of Science 43. ↵ Kraker , A. J. et al. Biochemical and cellular effects of c-Src kinase-selective pyrido[2, 3-d]pyrimidine tyrosine kinase inhibitors . Biochem Pharmacol 60 , 885 – 898 ( 2000 ). OpenUrl CrossRef PubMed Web of Science 44. ↵ Roche , S. , Fumagalli , S. & Courtneidge , S. A. Requirement for Src family protein tyrosine kinases in G2 for fibroblast cell division . Science 269 , 1567 – 1569 ( 1995 ). OpenUrl Abstract / FREE Full Text 45. ↵ Alexeyenko , A. & Sonnhammer , E. L. L. Global networks of functional coupling in eukaryotes from comprehensive data integration . Genome Res . 19 , 1107 – 1116 ( 2009 ). OpenUrl Abstract / FREE Full Text 46. Kim , S.-H. & Yi , S. V. Correlated asymmetry of sequence and functional divergence between duplicate proteins of Saccharomyces cerevisiae . Mol Biol Evol 23 , 1068 – 1075 ( 2006 ). OpenUrl CrossRef PubMed Web of Science 47. Konaté , M. M. et al. Molecular function limits divergent protein evolution on planetary timescales . Elife 8 , ( 2019 ). 48. Gong , S. et al. Structural and functional restraints in the evolution of protein families and superfamilies . Biochem Soc Trans 37 , 727 – 733 ( 2009 ). OpenUrl Abstract / FREE Full Text 49. ↵ Kaltenegger , E. & Ober , D. Paralogue Interference Affects the Dynamics after Gene Duplication . Trends Plant Sci . 20 , 814 – 821 ( 2015 ). OpenUrl CrossRef PubMed 50. ↵ Wente , S. R. & Schachman , H. K. Shared active sites in oligomeric enzymes: model studies with defective mutants of aspartate transcarbamoylase produced by site-directed mutagenesis . Proc Natl Acad Sci U S A 84 , 31 – 35 ( 1987 ). OpenUrl Abstract / FREE Full Text 51. Zwickl , P. , Kleinz , J. & Baumeister , W. Critical elements in proteasome assembly . Nat Struct Biol 1 , 765 – 770 ( 1994 ). OpenUrl CrossRef PubMed Web of Science 52. ↵ Ispolatov , I. , Yuryev , A. , Mazo , I. & Maslov , S. Binding properties and evolution of homodimers in protein-protein interaction networks . Nucleic Acids Res 33 , 3629 – 3635 ( 2005 ). OpenUrl CrossRef PubMed Web of Science 53. ↵ Pérez-Bercoff , A. , Makino , T. & McLysaght , A. Duplicability of self-interacting human genes . BMC Evol Biol 10 , 160 ( 2010 ). OpenUrl CrossRef PubMed 54. ↵ Pereira-Leal , J. B. , Levy , E. D. , Kamp , C. & Teichmann , S. A. Evolution of protein complexes by duplication of homomeric interactions . Genome Biol . 8 , R51 ( 2007 ). OpenUrl CrossRef PubMed 55. ↵ Mallik , S. & Tawfik , D. S. Determining the interaction status and evolutionary fate of duplicated homomeric proteins . PLoS Comput. Biol . 16 , e1008145 ( 2020 ). OpenUrl CrossRef PubMed 56. ↵ Mallik , S. , Tawfik , D. S. & Levy , E. D. How gene duplication diversifies the landscape of protein oligomeric state and function . Curr. Opin. Genet. Dev . 76 , 101966 ( 2022 ). OpenUrl CrossRef PubMed 57. ↵ Marchant , A. et al. The role of structural pleiotropy and regulatory evolution in the retention of heteromers of paralogs . Elife 8 , ( 2019 ). 58. ↵ Hochberg , G. K. A. et al. Structural principles that enable oligomeric small heat-shock protein paralogs to evolve distinct functions . Science 359 , 930 – 935 ( 2018 ). OpenUrl Abstract / FREE Full Text 59. ↵ Finnigan , G. C. , Hanson-Smith , V. , Stevens , T. H. & Thornton , J. W. Evolution of increased complexity in a molecular machine . Nature 481 , 360 – 364 ( 2012 ). OpenUrl CrossRef PubMed Web of Science 60. Pillai , A. S. et al. Origin of complexity in haemoglobin evolution . Nature 581 , 480 – 485 ( 2020 ). OpenUrl CrossRef PubMed 61. ↵ Després , P. C. et al. Compensatory mutations potentiate constructive neutral evolution by gene duplication . Science 385 , 770 – 775 ( 2024 ). OpenUrl CrossRef PubMed 62. ↵ Veretnik , S. , Wills , C. , Youkharibache , P. , Valas , R. E. & Bourne , P. E. Sm/Lsm genes provide a glimpse into the early evolution of the spliceosome . PLoS Comput. Biol . 5 , e1000315 ( 2009 ). OpenUrl CrossRef PubMed 63. Bochman , M. L. & Schwacha , A. The Mcm complex: unwinding the mechanism of a replicative helicase . Microbiol Mol Biol Rev 73 , 652 – 683 ( 2009 ). OpenUrl Abstract / FREE Full Text 64. Gupta , R. S. Evolution of the chaperonin families (Hsp60, Hsp10 and Tcp-1) of proteins and the origin of eukaryotic cells . Mol. Microbiol . 15 , 1 – 11 ( 1995 ). OpenUrl CrossRef PubMed Web of Science 65. ↵ Gille , C. et al. A comprehensive view on proteasomal sequences: implications for the evolution of the proteasome . J. Mol. Biol . 326 , 1437 – 1448 ( 2003 ). OpenUrl CrossRef PubMed Web of Science 66. ↵ Kumar , S. et al. TimeTree 5: An Expanded Resource for Species Divergence Times . Mol. Biol. Evol . 39 , ( 2022 ). 67. ↵ Altschul , S. F. , Gish , W. , Miller , W. , Myers , E. W. & Lipman , D. J. Basic local alignment search tool . J. Mol. Biol . 215 , 403 – 410 ( 1990 ). OpenUrl CrossRef PubMed Web of Science 68. ↵ Enright , A. J. , Van Dongen , S. & Ouzounis , C. A. An efficient algorithm for large-scale detection of protein families . Nucleic Acids Res 30 , 1575 – 1584 ( 2002 ). OpenUrl CrossRef PubMed Web of Science 69. ↵ Emms , D. M. & Kelly , S. OrthoFinder: phylogenetic orthology inference for comparative genomics . Genome Biol . 20 , 238 ( 2019 ). OpenUrl CrossRef PubMed 70. ↵ Berman , H. M. et al. The Protein Data Bank . Nucleic Acids Res . 28 , 235 – 242 ( 2000 ). OpenUrl CrossRef PubMed Web of Science 71. ↵ Schweke , H. et al. An atlas of protein homo-oligomerization across domains of life . Cell 187 , 999 – 1010 .e15 ( 2024 ). OpenUrl CrossRef PubMed 72. ↵ Meldal , B. H. M. et al. Complex Portal 2022: new curation frontiers . Nucleic Acids Res . 50 , D578 – D586 ( 2022 ). OpenUrl CrossRef PubMed 73. ↵ Drew , K. , Wallingford , J. B. & Marcotte , E. M. hu.MAP 2.0: integration of over 15,000 proteomic experiments builds a global compendium of human multiprotein assemblies . Mol. Syst. Biol . 17 , e10016 ( 2021 ). OpenUrl CrossRef PubMed 74. ↵ Stark , C. et al. BioGRID: a general repository for interaction datasets . Nucleic Acids Res . 34 , D535 – 9 ( 2006 ). OpenUrl CrossRef PubMed Web of Science 75. ↵ Orchard , S. et al. The MIntAct project--IntAct as a common curation platform for 11 molecular interaction databases . Nucleic Acids Res . 42 , D358 – 63 ( 2014 ). OpenUrl CrossRef PubMed Web of Science 76. ↵ Luck , K. et al. A reference map of the human binary protein interactome . Nature 580 , 402 – 408 ( 2020 ). OpenUrl CrossRef PubMed 77. ↵ Calderone , A. , Castagnoli , L. & Cesareni , G. mentha: a resource for browsing integrated protein-interaction networks . Nat. Methods 10 , 690 – 691 ( 2013 ). OpenUrl CrossRef PubMed Web of Science 78. ↵ Chatr-aryamontri , A. et al. MINT: the Molecular INTeraction database . Nucleic Acids Res . 35 , D572 – 4 ( 2007 ). OpenUrl CrossRef PubMed Web of Science 79. ↵ Das , J. & Yu , H. HINT: High-quality protein interactomes and their applications in understanding human disease . BMC Syst. Biol . 6 , 92 ( 2012 ). OpenUrl CrossRef PubMed 80. ↵ Alanis-Lobato , G. , Andrade-Navarro , M. A. & Schaefer , M. H. HIPPIE v2.0: enhancing meaningfulness and reliability of protein-protein interaction networks . Nucleic Acids Res . 45 , D408 – D414 ( 2017 ). OpenUrl CrossRef PubMed 81. ↵ Ishikawa , S. A. , Zhukova , A. , Iwasaki , W. & Gascuel , O. A Fast Likelihood Method to Reconstruct and Visualize Ancestral Scenarios . Mol. Biol. Evol . 36 , 2069 – 2085 ( 2019 ). OpenUrl CrossRef PubMed 82. ↵ Dehal , P. & Boore , J. L. Two rounds of whole genome duplication in the ancestral vertebrate . PLoS Biol . 3 , e314 ( 2005 ). OpenUrl CrossRef PubMed 83. Holland , L. Z. & Ocampo Daza , D. A new look at an old question: when did the second whole genome duplication occur in vertebrate evolution? Genome Biol . 19 , 209 ( 2018 ). OpenUrl CrossRef PubMed 84. Makino , T. & McLysaght , A. Ohnologs in the human genome are dosage balanced and frequently associated with disease . Proc. Natl. Acad. Sci. U. S. A . 107 , 9270 – 9274 ( 2010 ). OpenUrl Abstract / FREE Full Text 85. ↵ Yu , D. et al. Hagfish genome elucidates vertebrate whole-genome duplication events and their evolutionary consequences . Nat Ecol Evol 8 , 519 – 535 ( 2024 ). OpenUrl CrossRef PubMed 86. ↵ Vosseberg , J. et al. Timing the origin of eukaryotic cellular complexity with ancient duplications . Nat Ecol Evol 5 , 92 – 100 ( 2021 ). OpenUrl PubMed 87. Tria , F. D. K. et al. Gene Duplications Trace Mitochondria to the Onset of Eukaryote Complexity . Genome Biol. Evol . 13 , ( 2021 ). 88. Zhou , X. , Lin , Z. & Ma , H. Phylogenetic detection of numerous gene duplications shared by animals, fungi and plants . Genome Biol . 11 , R38 ( 2010 ). OpenUrl CrossRef PubMed 89. Makarova , K. S. , Wolf , Y. I. , Mekhedov , S. L. , Mirkin , B. G. & Koonin , E. V. Ancestral paralogs and pseudoparalogs and their role in the emergence of the eukaryotic cell . Nucleic Acids Res . 33 , 4626 – 4638 ( 2005 ). OpenUrl CrossRef PubMed Web of Science 90. Koonin , E. V. The origin and early evolution of eukaryotes in the light of phylogenomics . Genome Biol . 11 , 209 ( 2010 ). OpenUrl CrossRef PubMed 91. ↵ Roger , A. J. , Susko , E. & Leger , M. M. Evolution: Reconstructing the Timeline of Eukaryogenesis . Curr. Biol . 31 , R193 – R196 ( 2021 ). OpenUrl CrossRef PubMed 92. ↵ Fernández , R. & Gabaldón , T. Gene gain and loss across the metazoan tree of life . Nat Ecol Evol 4 , 524 – 533 ( 2020 ). OpenUrl CrossRef PubMed 93. Larroux , C. et al. Genesis and expansion of metazoan transcription factor gene classes . Mol. Biol. Evol . 25 , 980 – 996 ( 2008 ). OpenUrl CrossRef PubMed Web of Science 94. ↵ Lundin , L. G. Gene duplications in early metazoan evolution . Semin. Cell Dev. Biol . 10 , 523 – 530 ( 1999 ). OpenUrl CrossRef PubMed Web of Science 95. ↵ Kanehisa , M. & Goto , S. KEGG: kyoto encyclopedia of genes and genomes . Nucleic Acids Res . 28 , 27 – 30 ( 2000 ). OpenUrl CrossRef PubMed Web of Science 96. ↵ Cisneros , A. F. , Nielly-Thibault , L. , Mallik , S. , Levy , E. D. & Landry , C. R. Mutational biases favor complexity increases in protein interaction networks after gene duplication . Mol. Syst. Biol . 20 , 549 – 572 ( 2024 ). OpenUrl CrossRef PubMed 97. ↵ Kimura , M. Evolutionary rate at the molecular level . Nature 217 , 624 – 626 ( 1968 ). OpenUrl CrossRef PubMed Web of Science 98. ↵ Ota , T. & Kimura , M. On the constancy of the evolutionary rate of cistrons . J. Mol. Evol . 1 , 18 – 25 ( 1971 ). OpenUrl CrossRef PubMed 99. ↵ Delgado , J. , Radusky , L. G. , Cianferoni , D. & Serrano , L. FoldX 5.0: working with RNA, small molecules and a new graphical interface . Bioinformatics 35 , 4168 – 4169 ( 2019 ). OpenUrl CrossRef PubMed 100. ↵ Tiberti , M. et al. MutateX: an automated pipeline for in silico saturation mutagenesis of protein structures and structural ensembles . Brief. Bioinform . 23 , bbac074 ( 2022 ). OpenUrl CrossRef PubMed 101. ↵ Cheng , H. et al. ECOD: an evolutionary classification of protein domains . PLoS Comput. Biol . 10 , e1003926 ( 2014 ). OpenUrl CrossRef PubMed 102. ↵ Steinkellner , G. et al. Identification of promiscuous ene-reductase activity by mining structural databases using active site constellations . Nat. Commun . 5 , 4150 ( 2014 ). OpenUrl CrossRef PubMed 103. ↵ Hashimoto , K. & Panchenko , A. R. Mechanisms of protein oligomerization, the critical role of insertions and deletions in maintaining different oligomeric states . Proc. Natl. Acad. Sci. U. S. A . 107 , 20352 – 20357 ( 2010 ). OpenUrl Abstract / FREE Full Text 104. ↵ Akiva , E. , Itzhaki , Z. & Margalit , H. Built-in loops allow versatility in domain-domain interactions: lessons from self-interacting domains . Proc. Natl. Acad. Sci. U. S. A . 105 , 13292 – 13297 ( 2008 ). OpenUrl Abstract / FREE Full Text 105. ↵ Thul , P. J. & Lindskog , C. The human protein atlas: A spatial map of the human proteome . Protein Sci . 27 , 233 – 244 ( 2018 ). OpenUrl CrossRef PubMed 106. ↵ Jumper , J. et al. Highly accurate protein structure prediction with AlphaFold . Nature 596 , 583 – 589 ( 2021 ). OpenUrl CrossRef PubMed 107. ↵ Evans , R. et al. Protein complex prediction with AlphaFold-Multimer . bioRxiv 2021.10.04.463034 ( 2022 ) doi: 10.1101/2021.10.04.463034 . OpenUrl Abstract / FREE Full Text 108. ↵ Morales-Polanco , F. , Lee , J. H. , Barbosa , N. M. & Frydman , J. Cotranslational Mechanisms of Protein Biogenesis and Complex Assembly in Eukaryotes . Annu Rev Biomed Data Sci 5 , 67 – 94 ( 2022 ). OpenUrl CrossRef PubMed 109. Schwarz , A. & Beck , M. The Benefits of Cotranslational Assembly: A Structural Perspective . Trends Cell Biol . 29 , 791 – 803 ( 2019 ). OpenUrl CrossRef PubMed 110. ↵ Shiber , A. et al. Cotranslational assembly of protein complexes in eukaryotes revealed by ribosome profiling . Nature 561 , 268 – 272 ( 2018 ). OpenUrl CrossRef PubMed 111. ↵ Mallik , S. et al. Structural determinants of co-translational protein complex assembly . bioRxiv ( 2024 ) doi: 10.1101/2024.01.20.576408 . OpenUrl Abstract / FREE Full Text 112. ↵ Bertolini , M. et al. Interactions between nascent proteins translated by adjacent ribosomes drive homomer assembly . Science 371 , 57 – 64 ( 2021 ). OpenUrl Abstract / FREE Full Text 113. ↵ Mathieson , T. et al. Systematic analysis of protein turnover in primary cells . Nat. Commun . 9 , 689 ( 2018 ). OpenUrl CrossRef PubMed 114. Eden , E. et al. Proteome half-life dynamics in living human cells . Science 331 , 764 – 768 ( 2011 ). OpenUrl Abstract / FREE Full Text 115. ↵ Mallik , S. & Kundu , S. Topology and Oligomerization of Mono- and Oligomeric Proteins Regulate Their Half-Lives in the Cell . Structure 26 , 869 – 878 .e3 ( 2018 ). OpenUrl CrossRef 116. ↵ Bairoch , A. The ENZYME database in 2000 . Nucleic Acids Res . 28 , 304 – 305 ( 2000 ). OpenUrl CrossRef PubMed Web of Science 117. ↵ Carbon , S. et al. AmiGO: online access to ontology and annotation data . Bioinformatics 25 , 288 – 289 ( 2008 ). OpenUrl CrossRef PubMed 118. ↵ Binns , D. et al. QuickGO: a web-based tool for Gene Ontology searching . Bioinformatics 25 , 3045 – 3046 ( 2009 ). OpenUrl CrossRef PubMed Web of Science 119. ↵ Neuenfeldt , A. , Just , A. , Betat , H. & Mörl , M. Evolution of tRNA nucleotidyltransferases: a small deletion generated CC-adding enzymes . Proc. Natl. Acad. Sci. U. S. A . 105 , 7953 – 7958 ( 2008 ). OpenUrl Abstract / FREE Full Text 120. Afriat-Jurnou , L. , Jackson , C. J. & Tawfik , D. S. Reconstructing a missing link in the evolution of a recently diverged phosphotriesterase by active-site loop remodeling . Biochemistry 51 , 6047 – 6055 ( 2012 ). OpenUrl CrossRef PubMed Web of Science 121. Britten , R. J. Transposable element insertions have strongly affected human evolution . Proc. Natl. Acad. Sci. U. S. A . 107 , 19945 – 19948 ( 2010 ). OpenUrl Abstract / FREE Full Text 122. ↵ Tóth-Petróczy , A. & Tawfik , D. S. Protein insertions and deletions enabled by neutral roaming in sequence space . Mol. Biol. Evol . 30 , 761 – 771 ( 2013 ). OpenUrl CrossRef PubMed Web of Science 123. ↵ Zhang , Y. & Skolnick , J. Scoring function for automated assessment of protein structure template quality . Proteins 57 , 702 – 710 ( 2004 ). OpenUrl CrossRef PubMed Web of Science 124. ↵ Goldman , N. & Yang , Z. A codon-based model of nucleotide substitution for protein-coding DNA sequences . Mol. Biol. Evol . 11 , 725 – 736 ( 1994 ). OpenUrl CrossRef PubMed Web of Science 125. ↵ Muse , S. V. & Gaut , B. S. A likelihood approach for comparing synonymous and nonsynonymous nucleotide substitution rates, with application to the chloroplast genome . Mol. Biol. Evol . 11 , 715 – 724 ( 1994 ). OpenUrl CrossRef PubMed Web of Science 126. ↵ Yang , Z. , Nielsen , R. , Goldman , N. & Pedersen , A. M. Codon-substitution models for heterogeneous selection pressure at amino acid sites . Genetics 155 , 431 – 449 ( 2000 ). OpenUrl Abstract / FREE Full Text 127. Nielsen , R. & Yang , Z. Likelihood models for detecting positively selected amino acid sites and applications to the HIV-1 envelope gene . Genetics 148 , 929 – 936 ( 1998 ). OpenUrl Abstract / FREE Full Text 128. ↵ Kryazhimskiy , S. & Plotkin , J. B. The population genetics of dN/dS . PLoS Genet . 4 , e1000304 ( 2008 ). OpenUrl CrossRef PubMed 129. ↵ Latrille , T. , Rodrigue , N. & Lartillot , N. Genes and sites under adaptation at the phylogenetic scale also exhibit adaptation at the population-genetic scale . Proc. Natl. Acad. Sci. U. S. A . 120 , e2214977120 ( 2023 ). OpenUrl CrossRef PubMed 130. ↵ O’Leary , M. A. et al. The placental mammal ancestor and the post-K-Pg radiation of placentals . Science 339 , 662 – 667 ( 2013 ). OpenUrl Abstract / FREE Full Text 131. ↵ Hochberg , G. K. A. et al. A hydrophobic ratchet entrenches molecular complexes . Nature 588 , 503 – 508 ( 2020 ). OpenUrl CrossRef PubMed 132. Emlaw , J. R. et al. A single historical substitution drives an increase in acetylcholine receptor complexity . Proc. Natl. Acad. Sci. U. S. A . 118 , ( 2021 ). 133. ↵ Stutz , C. & Blein , S. A single mutation increases heavy-chain heterodimer assembly of bispecific antibodies by inducing structural disorder in one homodimer species . J. Biol. Chem . 295 , 9392 – 9408 ( 2020 ). OpenUrl Abstract / FREE Full Text 134. ↵ Stoltzfus , A. On the possibility of constructive neutral evolution . J. Mol. Evol . 49 , 169 – 181 ( 1999 ). OpenUrl CrossRef PubMed Web of Science 135. ↵ Lynch , M. Evolution of the mutation rate . Trends Genet . 26 , 345 – 352 ( 2010 ). OpenUrl CrossRef PubMed Web of Science 136. ↵ Sung , W. , Ackerman , M. S. , Miller , S. F. , Doak , T. G. & Lynch , M. Drift-barrier hypothesis and mutation-rate evolution . Proc. Natl. Acad. Sci. U. S. A . 109 , 18488 – 18492 ( 2012 ). OpenUrl Abstract / FREE Full Text 137. ↵ Badonyi , M. & Marsh , J. A. Buffering of genetic dominance by allele-specific protein complex assembly . Sci Adv 9 , eadf9845 ( 2023 ). OpenUrl CrossRef PubMed 138. ↵ Natan , E. , Wells , J. N. , Teichmann , S. A. & Marsh , J. A. Regulation, evolution and consequences of cotranslational protein complex assembly . Curr. Opin. Struct. Biol . 42 , 90 – 97 ( 2017 ). OpenUrl CrossRef PubMed 139. ↵ Conant , G. C. & Wagner , A. Asymmetric sequence divergence of duplicate genes . Genome Res . 13 , 2052 – 2058 ( 2003 ). OpenUrl Abstract / FREE Full Text 140. Panchin , A. Y. , Gelfand , M. S. , Ramensky , V. E. & Artamonova , I. I. Asymmetric and non-uniform evolution of recently duplicated human genes . Biol. Direct 5 , 54 ( 2010 ). OpenUrl CrossRef PubMed 141. Yang , J. , Xie , Z. & Glover , B. J. Asymmetric evolution of duplicate genes encoding the CCAAT-binding factor NF-Y in plant genomes . New Phytol . 165 , 623 – 631 ( 2005 ). OpenUrl CrossRef PubMed Web of Science 142. Holland , P. W. H. , Marlétaz , F. , Maeso , I. , Dunwell , T. L. & Paps , J. New genes from old: asymmetric divergence of gene duplicates and the evolution of development . Philos. Trans. R. Soc. Lond. B Biol. Sci . 372 , ( 2017 ). 143. Cusack , B. P. & Wolfe , K. H. Not born equal: increased rate asymmetry in relocated and retrotransposed rodent gene duplicates . Mol. Biol. Evol . 24 , 679 – 686 ( 2007 ). OpenUrl CrossRef PubMed Web of Science 144. ↵ Pegueroles , C. , Laurie , S. & Albà , M. M. Accelerated evolution after gene duplication: a time-dependent process affecting just one copy . Mol. Biol. Evol . 30 , 1830 – 1842 ( 2013 ). OpenUrl CrossRef PubMed Web of Science 145. ↵ Rebeaud , M. E. , Mallik , S. , Goloubinoff , P. & Tawfik , D. S. On the evolution of chaperones and cochaperones and the expansion of proteomes across the Tree of Life . Proc. Natl. Acad. Sci. U. S. A . 118 , ( 2021 ). 146. ↵ Zaremba-Niedzwiedzka , K. et al. Asgard archaea illuminate the origin of eukaryotic cellular complexity . Nature 541 , 353 – 358 ( 2017 ). OpenUrl CrossRef PubMed 147. ↵ The UniProt Consortium et al. UniProt: the Universal Protein Knowledgebase in 2023 . Nucleic Acids Res . 51 , D523 – D531 ( 2022 ). OpenUrl 148. ↵ Yang , Y. & Smith , S. A. Orthology inference in nonmodel organisms using transcriptomes and low-coverage genomes: improving accuracy and matrix occupancy for phylogenomics . Mol. Biol. Evol . 31 , 3081 – 3092 ( 2014 ). OpenUrl CrossRef PubMed 149. ↵ Griesmann , M. et al. Phylogenomics reveals multiple losses of nitrogen-fixing root nodule symbiosis . Science 361 , ( 2018 ). 150. ↵ Di Tommaso , P. et al. T-Coffee: a web server for the multiple sequence alignment of protein and RNA sequences using structural information and homology extension . Nucleic Acids Res . 39 , W13 – 7 ( 2011 ). OpenUrl CrossRef PubMed Web of Science 151. ↵ Varadi , M. et al. AlphaFold Protein Structure Database in 2024: providing structure coverage for over 214 million protein sequences . Nucleic Acids Res . 52 , D368 – D375 ( 2024 ). OpenUrl CrossRef PubMed 152. ↵ Capella-Gutiérrez , S. , Silla-Martínez , J. M. & Gabaldón , T. trimAl: a tool for automated alignment trimming in large-scale phylogenetic analyses . Bioinformatics 25 , 1972 – 1973 ( 2009 ). OpenUrl CrossRef PubMed Web of Science 153. ↵ Dey , S. , Ritchie , D. W. & Levy , E. D. PDB-wide identification of biological assemblies from conserved quaternary structure geometry . Nat. Methods 15 , 67 – 72 ( 2018 ). OpenUrl CrossRef PubMed 154. ↵ Levy , E. D. , Pereira-Leal , J. B. , Chothia , C. & Teichmann , S. A. 3D complex: a structural classification of protein complexes . PLoS Comput. Biol . 2 , e155 ( 2006 ). OpenUrl CrossRef PubMed 155. ↵ Chen , K. , Durand , D. & Farach-Colton , M. NOTUNG: a program for dating gene duplications and optimizing gene family trees . J. Comput. Biol . 7 , 429 – 447 ( 2000 ). OpenUrl CrossRef PubMed Web of Science 156. ↵ Li , W. & Godzik , A. Cd-hit: a fast program for clustering and comparing large sets of protein or nucleotide sequences . Bioinformatics 22 , 1658 – 1659 ( 2006 ). OpenUrl CrossRef PubMed Web of Science 157. ↵ Fu , L. , Niu , B. , Zhu , Z. , Wu , S. & Li , W. CD-HIT: accelerated for clustering the next-generation sequencing data . Bioinformatics 28 , 3150 – 3152 ( 2012 ). OpenUrl CrossRef PubMed Web of Science 158. ↵ Usmanova , D. R. et al. Self-consistency test reveals systematic bias in programs for prediction change of stability upon mutation . Bioinformatics 34 , 3653 – 3658 ( 2018 ). OpenUrl CrossRef PubMed 159. ↵ Pace , C. N. The stability of globular proteins . CRC Crit. Rev. Biochem . 3 , 1 – 43 ( 1975 ). OpenUrl CrossRef PubMed 160. ↵ Plaxco , K. W. , Simons , K. T. , Ruczinski , I. & Baker , D. Topology, stability, sequence, and length: defining the determinants of two-state protein folding kinetics . Biochemistry 39 , 11177 – 11183 ( 2000 ). OpenUrl CrossRef PubMed Web of Science 161. ↵ Choi , J.-M. et al. Minimalistic predictor of protein binding energy: contribution of solvation factor to protein binding . Biophys. J . 108 , 795 – 798 ( 2015 ). OpenUrl CrossRef PubMed 162. ↵ Jankauskaite , J. , Jiménez-García , B. , Dapkūnas , J. , Fernández-Recio , J. & Moal , I. H. SKEMPI 2.0: an updated benchmark of changes in protein–protein binding energy, kinetics and thermodynamics upon mutation . Bioinformatics 35 , 462 – 469 ( 2018 ). OpenUrl 163. ↵ Sailer , Z. R. & Harms , M. J. Molecular ensembles make evolution unpredictable . Proc. Natl. Acad. Sci. U. S. A . 114 , 11938 – 11943 ( 2017 ). OpenUrl Abstract / FREE Full Text 164. ↵ Mirdita , M. et al. ColabFold: making protein folding accessible to all . Nat. Methods 19 , 679 – 682 ( 2022 ). OpenUrl CrossRef PubMed 165. ↵ Steinegger , M. & Söding , J. MMseqs2 enables sensitive protein sequence searching for the analysis of massive data sets . Nat. Biotechnol . 35 , 1026 – 1028 ( 2017 ). OpenUrl CrossRef PubMed 166. ↵ Edgar , R. C. Muscle5: High-accuracy alignment ensembles enable unbiased assessments of sequence homology and phylogeny . Nat. Commun . 13 , 1 – 9 ( 2022 ). OpenUrl CrossRef PubMed 167. ↵ Mukherjee , S. & Zhang , Y. MM-align: a quick algorithm for aligning multiple-chain protein complex structures using iterative dynamic programming . Nucleic Acids Res . 37 , e83 – e83 ( 2009 ). OpenUrl CrossRef PubMed 168. ↵ Basu , S. & Wallner , B. DockQ: A quality measure for protein-protein docking models . PLoS One 11 , e0161879 ( 2016 ). OpenUrl CrossRef PubMed 169. ↵ Abramson , J. et al. Accurate structure prediction of biomolecular interactions with AlphaFold 3 . Nature 630 , 493 – 500 ( 2024 ). OpenUrl CrossRef PubMed 170. ↵ Levy , E. D. A simple definition of structural regions in proteins and its use in analyzing interface evolution . J. Mol. Biol . 403 , 660 – 670 ( 2010 ). OpenUrl CrossRef PubMed Web of Science 171. ↵ Mitternacht , S. FreeSASA: An open source C library for solvent accessible surface area calculations . F1000Res . 5 , 189 ( 2016 ). OpenUrl 172. ↵ Rice , P. , Longden , I. & Bleasby , A. EMBOSS: the European Molecular Biology Open Software Suite . Trends Genet . 16 , 276 – 277 ( 2000 ). OpenUrl CrossRef PubMed Web of Science 173. ↵ Sievers , F. & Higgins , D. G. Clustal Omega for making accurate alignments of many protein sequences . Protein Sci 27 , 135 – 145 ( 2018 ). OpenUrl CrossRef PubMed 174. ↵ Huntley , R. P. et al. The GOA database: gene Ontology annotation updates for 2015 . Nucleic Acids Res . 43 , D1057 – 63 ( 2015 ). OpenUrl CrossRef PubMed 175. ↵ Rodrigue , N. , Latrille , T. & Lartillot , N. A Bayesian Mutation-Selection Framework for Detecting Site-Specific Adaptive Evolution in Protein-Coding Genes . Mol. Biol. Evol . 38 , 1199 – 1208 ( 2021 ). OpenUrl CrossRef PubMed View the discussion thread. Back to top Previous Next Posted January 23, 2025. Download PDF Supplementary Material Data/Code 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 Co-translational assembly promotes functional diversification of paralogous 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 Co-translational assembly promotes functional diversification of paralogous proteins Saurav Mallik , Angel F. Cisneros , Christian R. Landry , Emmanuel D. Levy bioRxiv 2025.01.22.634331; doi: https://doi.org/10.1101/2025.01.22.634331 Share This Article: Copy Citation Tools Co-translational assembly promotes functional diversification of paralogous proteins Saurav Mallik , Angel F. Cisneros , Christian R. Landry , Emmanuel D. Levy bioRxiv 2025.01.22.634331; doi: https://doi.org/10.1101/2025.01.22.634331 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 Evolutionary Biology Subject Areas All Articles Animal Behavior and Cognition (7628) Biochemistry (17659) Bioengineering (13881) Bioinformatics (41909) Biophysics (21435) Cancer Biology (18576) Cell Biology (25478) Clinical Trials (138) Developmental Biology (13366) Ecology (19885) Epidemiology (2067) Evolutionary Biology (24299) Genetics (15597) Genomics (22482) Immunology (17725) Microbiology (40356) Molecular Biology (17160) Neuroscience (88523) Paleontology (666) Pathology (2830) Pharmacology and Toxicology (4820) Physiology (7636) Plant Biology (15123) Scientific Communication and Education (2044) Synthetic Biology (4290) Systems Biology (9817) Zoology (2269)