Unicore Enables Scalable and Accurate Phylogenetic Reconstruction with Structural Core Genes

Unicore Enables Scalable and Accurate Phylogenetic Reconstruction with Structural Core Genes | 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 Unicore Enables Scalable and Accurate Phylogenetic Reconstruction with Structural Core Genes View ORCID Profile Dongwook Kim , View ORCID Profile Sukhwan Park , View ORCID Profile Martin Steinegger doi: https://doi.org/10.1101/2024.12.22.629535 Dongwook Kim 1 Interdisciplinary Program in Bioinformatics, Seoul National University , Seoul 08826, Republic of Korea Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Dongwook Kim Sukhwan Park 1 Interdisciplinary Program in Bioinformatics, Seoul National University , Seoul 08826, Republic of Korea Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Sukhwan Park Martin Steinegger 1 Interdisciplinary Program in Bioinformatics, Seoul National University , Seoul 08826, Republic of Korea 2 School of Biological Sciences, Seoul National University , Seoul 08826, Republic of Korea 3 Institute of Molecular Biology and Genetics , Seoul 08826, Republic of Korea 4 Artificial Intelligence Institute, Seoul National University , Seoul 08826, Republic of Korea Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Martin Steinegger For correspondence: martin.steinegger{at}snu.ac.kr Abstract Full Text Info/History Metrics Supplementary material Preview PDF Abstract The analysis of single-copy core genes, common to most members of a clade, is important for key tasks in biology including phylogenetic reconstruction and assessing genome quality. Core genes are traditionally identified by the analysis of amino acid similarities among proteomes, but can also be defined using structures, which bear potential in deep clades beyond the twilight zone of amino acids. Despite breakthroughs in accurate AI-driven protein structure prediction, obtaining full 3D structural models on a proteomic scale is still prohibitively slow. Here, we present Unicore, a novel method for identifying structural core genes at a scale suitable for downstream phylogenetic analysis. By applying the ProstT5 protein language model to the input proteomes to obtain their 3Di structural strings, Unicore saves over three orders of magnitude in runtime compared to a full 3D prediction. Using Foldseek clustering, Unicore identifies single-copy structures universally present in the species and aligns them using Foldmason. These structural core gene alignments are projected back to amino acid information for downstream phylogenetic analysis. We demonstrate that this approach defines core genes with linear run-time scaling over the number of species, up to 56 times faster than OrthoFinder, while reconstructing phylogenetic relationships congruent with conventional approaches. Unicore is universally applicable to any given set of taxa, even spanning superkingdoms and overcoming limitations of previous methods requiring orthologs of fixed taxonomic scope, and is available as a free and open source software at https://github.com/steineggerlab/unicore . Introduction Understanding the evolutionary relationships among organisms through phylogenetic analysis is a fundamental goal in biology. Phylogenomics allows us to reconstruct the tree of life, shedding light on how different species are related and how they have diverged over time ( 1 ). A common phylogenomic approach leverages single-copy core genes—universally conserved across taxa—as reliable markers for reconstructing phylogenetic relationships ( 2 ). State-of-the-art core gene databases, including OrthoDB ( 3 ), OMA ( 4 ), UBCG ( 5 ) or UFCG ( 6 ), offer lists of single-copy core genes for specific clades. These gene sets are detected from genomes through sequence-based orthology inference ( 7 ) or profile hidden Markov model (HMM)-based homology searches ( 8 ). However, HMM- or all-vs-all search-based ( 9 ) approaches are computationally intensive and less effective for newly discovered clades, particularly when they are in the ‘twilight zone’ from the known clades, where amino acid similarity is too low to detect homology ( 10 , 11 ). Protein structure-based methods, like Fold-seek ( 12 ) using the 3Di alphabet, overcome these challenges by detecting homologs below the twilight zone through structural searches ( 11 , 13 , 14 ). Furthermore, Foldseek’s linear-scaling time clustering enables rapid proteome-wide comparisons ( 15 , 16 ). The advent of AlphaFold2 ( 17 ) has opened the door for much broader use of structural information in phylogenetic analysis by lifting the dependency on limitedly available experimentally determined structures ( 18 , 19 ). Recent research shows that Foldseek’s 3Di alignments reveal deep evolutionary relationships and robust phylogenetic signals even for highly divergent sequences ( 20 – 22 ). However, applying AlphaFold2 or similar methods to even dozens of full proteomes for core gene detection would require months, limiting its use. Alternatively, protein language models (pLMs; ( 23 – 25 )) can be used to extract structural features directly from primary amino acid sequences, eliminating the need for computationally intensive structure prediction methods. Among them, ProstT5 ( 26 ) stands out, which is a bilingual pLM trained on millions of sequence-structure pairs from the AFDB ( 27 ). ProstT5 rapidly converts amino acid sequences into 3Di strings 3,600-fold faster than AlphaFold2-ColabFold ( 28 ), thereby making structure-informed large-scale phylogenetic analysis feasible. Here we present Unicore, a scalable and accurate method for structure-based core gene definition and phylogenetic inference. Unicore leverages predicted 3Di sequences from the ProstT5 model and linear-time comparison methods ( 16 ) to accelerate large-scale proteome analysis. It identifies single-copy ‘structural core genes’ on the fly without relying on pre-computed gene sets ( Fig. 1 ). By integrating Foldseek’s rapid structural comparisons, Unicore identifies conserved structural genes across proteomes, generates structural alignments from 3Di strings using FoldMason ( 29 ), and infers phylogenetic trees with maximum likelihood methods, such as IQ-TREE ( 30 ), FastTree ( 31 ), and RAxML ( 32 ). Download figure Open in new tab Fig. 1. Graphical illustration of the Unicore workflow. (a) The input species proteome, represented as amino acid sequences, is translated to 3Di alphabets (denoted by underscores) using the ProstT5 language model. These 3Di sequences are clustered with Foldseek to group structurally homologous proteins. Structural core genes are identified from these clusters by selecting those conserved as a single-copy in more than a specified proportion of the input species. (b) Foldmason is used to construct multiple structural alignments (MSTAs) for each structural core gene cluster. These alignments are then converted back into amino acid sequences, enabling conventional evolutionary model-based maximum likelihood phylogenetic inference. Finally, a species phylogenetic tree is generated from the concatenated MSAs. Methods Unicore is developed in a modular fashion with Rust programming language. Its modules: createdb , cluster , profile , and tree are described in the following sections. For ease of use and automated tree inference, easy-core orchestrates these modules into a single workflow. Defining structural core genes from species proteomes Unicore’s first module createdb takes a collection of species as an input, where each species is represented as a FASTA file of its proteome ( Fig. 1a ). These are translated into Foldseek’s 3Di alphabet ( 12 ) through the ProstT5 encoder-CNN ( 26 ). Both the 3Di sequences and their corresponding amino acid sequences are compiled together into a Foldseek database. The database is then clustered with the Foldseek cluster module, resulting in structural gene clusters. For each cluster, the module profile computes the fraction of species out of all input species that have exactly one member structure in that cluster. We denote this fraction as the single-copy coverage of the cluster. Clusters with coverage exceeding a certain threshold (in this study 80% was used as a default value), are defined as the structural core genes. Multiple structural alignment and phylogenetic inference From each structural core gene cluster, the module tree converts single-copy members into a Foldmason ( 29 ) multiple structural alignment (MSTA; Fig. 1b ). Gappy columns, defined as those with a gap fraction of 50% or higher, are then removed from the alignment. Finally, the remaining columns are converted back to amino acid sequences by mapping each 3Di alphabet to its corresponding residue, resulting in multiple sequence alignments (MSAs). From these structurally-aligned MSAs, Unicore reconstructs the phylogeny by running either IQ-TREE, FastTree, or RAxML ( 30 – 32 ). Unicore produces the maximum likelihood tree inferred from the concatenated MSAs of all identified structural core genes, as well as individual gene trees. Tree of life-wide phylogeny using Unicore We gathered the reference proteome from the UniProt release 2024_02 ( 33 ), filtered those with >1,000 proteins and containing >90% of single-copy BUSCOs (v5.4.7) with respective domain used as an OrthoDB v10 lineage ( 3 ), resulting in 9,163 proteomes. We then randomly sampled one species from each class, resulting in 166 species comprising all of the Bacterial, Archaeal, and Eukaryotic domains (Supplementary Table 1). We ran Unicore v0-eb516a3 easy-core module on these species with default parameters to produce a structural core gene tree with Foldseek v8-ef4e960, Foldmason v1-763a428, and IQ-TREE v2.2.2.3. Consistency benchmark - preparation of species sets Starting with the 9,163 quality-controlled proteomes from the tree of life analysis, we identified 188 taxa ranging from genus to phylum as having at least N member species (using N = 30 for bacteria and N = 20 for fungi), according to NCBI taxonomy annotations ( 34 ). We randomly divided each of these 188 taxa into sets of proteomes with exactly N members (e.g., fungal taxa with 41 species would yield two sets of 20 and one left out proteome), resulting in 876 bacterial and 98 fungal species sets (Supplementary Table 2). Of these, we randomly sampled one set from each bacterial taxa, resulting in 163 bacterial sets, which were used for the following analyses along with the entirety of 98 fungal sets (Supplementary Table 3). Consistency benchmark - comparison with sequence-based methods We computed structure- and sequence-based core genes from each of the 163 and 98 proteome sets using Unicore easy-core and OrthoFinder (v2.5.5), represented as single-copy genes present in at least 90% of the species within each set. This is more stringent than the default threshold, reducing the computational burden in the analysis of sets from lower taxonomic ranks (e.g., genus), where higher fractions of the proteomes are shared among set members. To this end, only the first three modules of Unicore were run with default parameters, except for profile , which was run with the option -t 0.9 . For OrthoFinder, we used the arguments -og -S mmseqs and extracted orthogroups that cover at least 90% of species within the dataset.Additionally, BUSCO v5.4.7 with default parameters was used to extract the respective lineage of OrthoDB v10 core genes from each set. For the core genes defined by each method, we performed HMM-HMM comparisons using hhsearch in HH-suite3 v3.3.0 ( 35 ) with MAFFT-generated MSAs as input ( 36 ). Each MSA was compared against core gene sets from other methods, with a pair classified as shared if the global alignment yielded an E -value < 10 − 3 with at least 80% coverage. Based on these comparisons, each core gene was classified as shared by both methods or as unique to a specific method. Assessment on phylogenetic robustness We utilized the Unicore, OrthoFinder, and BUSCO core genes of the benchmark species sets from the ‘Consistency benchmark’ section to compare the phylogenetic trees reconstructed from each tool. We used the tree module with default parameters to generate phylogenetic trees from the concatenated alignments of Unicore core genes. For the core genes of OrthoFinder and BUSCO, we computed MSAs using MAFFT-linsi v7.525 ( 36 ) with default settings, filtered out columns with over 50% gaps, and constructed phylogenetic trees using IQ-TREE v2.3.6 ( 30 ) with parameters -m JTT+F+I+G -B 1000 on the concatenated filtered MSAs. All resulting phylogenetic trees were rooted to yield minimum ancestor deviation ( 37 ). First, we calculated quartet similarity between the trees with tqDist v1.0.2 ( 38 , 39 ). Then we used the metrics introduced by FoldTree ( 22 ), namely taxonomic congruence scores (TCS) and ultrametricity, to further benchmark the quality of the phylogenetic trees. To test for statistical significance, we performed a t -test with the alternative hypothesis that Unicore trees exhibit either higher TCS or lower ultrametricity compared to those generated by OrthoFinder and BUSCO. We controlled for multiple tests by adjusting the p -values using the Benjamini-Hochberg method ( 40 ) for comparisons with OrthoFinder and BUSCO independently. Computation time benchmark We compared Unicore’s and OrthoFinder’s runtimes on sets of 10, 20, 30, … , 100 species. The species in these sets were randomly sampled from the 330 species of the 11 bacterial sets at the level ‘phyla’ identified as described in the ‘Consistency benchmark’ section (Supplementary Table 3). For Unicore, we used createdb with -max-len 4000 -g to generate 3Di databases, cluster with the arguments -c “-c 0.8 -max-seqs “ , where maxseqs is set to the larger of 1,000 or 20 times the number of species in the proteome database, and profile with default parameters. We separately estimated the time spent on 3Di conversion ( createdb ) and core gene definition ( cluster + profile ). Orthofinder was executed with the arguments -og -S mmseqs . We used one NVIDIA GeForce RTX 4090 to run Unicore createdb , and 128 threads of AMD EPYC 7702P to define core genes using both tools, and took an average runtime of 10 repeated runs. For both tools, we fit a linear regression model between the number of species and the tool’s runtime using R v4.4.1. Results and Discussion Phylogenetic reconstruction with structurally conserved core genes We identified 12 single-copy structural core genes with Unicore’s easy-core module (Supplementary Table 4), shared by at least 80% out of 166 species spanning the three domains of life (see Methods and Supplementary Table 1). Phylogenetic inference using these structural core genes demonstrated monophyletic delineation of each of the three domains, supported by a bootstrap average of 91.47 ( Fig. 2a , see Supplementary Fig. 1 for the detailed tree including species labels). An example of the structural conservation that characterizes Unicore’s core genes is the protein translocase subunit SecY (see Supplementary Table 4 for details), which was not identified as a core gene by OrthoFinder. SecY’s structures are conserved across Archaea, Bacteria, and Eukarya ( Fig. 2b ) and in each case, predicted 3Di string aligns well with the the experimentally defined structure (TM-score > 0.5). Download figure Open in new tab Fig. 2. Structural core genes resolve phylogenetic relationships between species with distantly homologous proteins. (a) Phylogenetic tree constructed from 12 structural core genes identified across 166 species, spanning Bacteria (pink), Archaea (blue), and Eukarya (green). Branches are color-coded by domain origin, while internal nodes are colored according to their bootstrap support values. (b) Structural conservation illustrated by SecY protein translocase, one of the identified core genes. Predicted structural models are shown for four species, highlighting structural homology across Bacteria ( Campylobacter hominis ), Archaea ( Methanococcus maripaludis ), Fungi ( Malassezia sympodialis ), and Animalia ( Capitella teleta ), annotated with the TM-score from the alignment against PDB entry 1rhz chain A. (c) Scatter plot comparing E -values derived from MMseqs2 and Foldseek alignments of sequentially distant protein pairs (<30% sequence identity). The data points represent an all-against-all comparison from six structural core genes uniquely defined by Unicore, which were not detected by OrthoFinder in the same species set. The gray dotted line indicates the equivalence E (MMseqs2) = E (Foldseek), dividing data into two partitions. Counts and proportions for each partition are reported. Next, we assessed the utility of structural comparison beyond the twilight zone by comparing E -values between MM-seqs2’s amino-acid and Foldseek’s 3Di alignments of structural core genes ( Fig. 2c ). We found that the majority (85.52%) of distantly homologous pairs were better supported (lower E -values) by Foldseek, underpinning the capability of 3Di alignments in elucidating distant homology for phylogenetics ( 20 – 22 ). Here, we applied an 80% threshold for single-copy species coverage to ensure a sufficient number of genes are found to conduct downstream phylogenetic analyses. Adjusting this coverage threshold can impact the resulting set of core genes, which inevitably affects the subsequent phylogenetic analysis ( 6 ). Therefore, we made this threshold customizable in Unicore, along with the parameters for Foldseek’s modules, allowing users flexibility in adjusting them according to their specific needs. Structural core genes are consistent with sequence-based methods We evaluated Unicore’s core genes by comparing them with those identified by sequence-based methods, specifically OrthoFinder ( 9 ) and BUSCO ( 41 ). Out of Unicore’s 170,638 core genes identified in 163 bacterial and 98 fungal sets (see Methods and Supplementary Table 3), 144,658 (84.75%) were detected as such by both OrthoFinder and BUSCO (Supplementary Fig. 2a), while 3,142 (1.84%) genes were unique to Unicore. In the opposite direction, we found 90.53% of OrthoFinder’s genes and 91.50% of BUSCO’s genes were detectable by Unicore (Supplementary Fig. 2b and 2c), further supporting the validity of Unicore’s core genes. Robustness of the phylogenetic trees reconstructed with Unicore We further assessed the quality of phylogenetic inference with Unicore structural core genes identified in the 163 bacterial and 98 fungal benchmark sets. Unicore’s bacterial trees showed median quartet similarity to OrthoFinder’s and BUSCO’s trees was 0.985 and 0.968, respectively ( Fig. 3a , left). This measure was even higher in the fungal dataset, reaching 1.0 compared to both methods ( Fig. 3a , right). However, this may result from having smaller fungal sets than bacterial sets (see Methods), an artifact due to the limited availability of high-quality fungal genomes. Download figure Open in new tab Fig. 3. Benchmark results comparing the speed and phylogenetic robustness of trees generated by Unicore, OrthoFinder, and BUSCO. (a) The histogram shows quartet similarity scores for trees computed with Unicore against OrthoFinder (yellow) and BUSCO (blue). The left panel displays bacterial trees, while the right shows fungal trees. A bin width of 0.005 is used, with the median marked by a vertical dotted line. (b) TCS and ultrametricity for Unicore (red), OrthoFinder (dark gray), and BUSCO (gray) are compared across taxonomic ranks. The Benjamini-Hochberg method corrected p-values, with significant results (p < 0.05) indicated by a star. (c) Left: Unicore (purple) includes the entire process from predicting 3Di sequences to core gene inference. Unicore with lookup (cyan) accounts only for predicting 3Di sequences absent in the AlphaFold database, while Unicore excl. createdb (red) benchmarks core gene inference alone. OrthoFinder (dark gray) measures orthogroup inference. Right: Results are plotted on a linear y-axis. Core gene inference with Unicore exhibits linear time complexity, unlike OrthoFinder’s quadratic scaling, as shown by the dotted trend line. Regression analysis confirmed this, with Unicore fitting poorly to a quadratic model (coefficient = -0.0002, p = 0.572) but strongly to a linear model (coefficient = 0.4932, p < 10 − 15 ). In contrast, OrthoFinder’s runtime fits well with a quadratic model (coefficient = 0.3095, p < 10 − 15 ). Meanwhile, Unicore trees showed comparable taxonomic congruence and ultrametricity with the trees computed based on the core genes of OrthoFinder and BUSCO, across all taxonomic ranks for both bacteria and fungi ( Fig. 3b ). Put together, structural core genes of Unicore reconstructed robust species phylogeny comparable to the sequence-only methods, highlighting both the validity of the structure-based phylogenetic approach and its prospect of elucidating beyond the twilight zone of amino acid sequences. Unicore allows structure-based phylogeny at scale To demonstrate the scalability of our method, we compared the time it takes Unicore and OrthoFinder to detect core genes in benchmark sets of species. Excluding database creation, Unicore scales linearly with the number of input species, thereby being faster than the quadratically-scaling OrthoFinder, reaching a factor of 58.69 in runtime over 100 species ( Fig. 3c ). However, database creation is Unicore’s biggest bottleneck, which requires 3Di conversion by ProstT5, making it up to 4.05 times slower than OrthoFinder. To alleviate this, we implemented a sequence lookup method that rapidly retrieves 3Di sequences from the AFDB ( 27 ) for exact sequence matches, thereby improving Unicore’s runtime by up to 5.44-fold, which was 1.34-fold faster than Orthofinder. Since conversion with ProstT5 heavily depends on the model and amount of GPU resources, the overall throughput is expected to be improved concurrent to the rapid development of graphics hardware. In this context, owing to its quasi-linear clustering algorithm ( 16 ), Unicore’s scalability will become increasingly powerful against the ever-increasing flood of biological data. Concluding remarks Unicore implements a large-scale, structure-based phylogenetic method powered by state-of-the-art tools for protein structural analysis, compiled into a fully automated bioinformatics pipeline. Accompanied by the recent developments in structural phylogenetics ( 20 – 22 , 29 , 42 ), we anticipate our tool to further expand the utility of protein structures from an evolutionary perspective. Data Availability Unicore is implemented in Rust and is available as GPLv3 licensed free open-source software at https://github.com/steineggerlab/unicore . Conflict of interest statement M.S. declares an outside interest in Stylus Medicine. Acknowledgments M.S. acknowledges support by the National Research Foundation of Korea grants (2020M3-A9G7-103933, 2021-R1C1-C102065 and 2021-M3A9-I4021220, RS-2024-00396026), Samsung DS research fund, Creative-Pioneering Researchers Program and AI-Bio Research Grant through Seoul National University. We thank Eli Levy Karin of ELKMO, as well as Rachel Seongeun Kim, Hyunbin Kim, and Jaebeom Kim of the Steinegger Lab for their help with revising the manuscript. References 1. ↵ Jonathan A Eisen and Claire M Fraser . Phylogenomics: intersection of evolution and genomics . Science , 300 ( 5626 ): 1706 – 1707 , 2003 . OpenUrl Abstract / FREE Full Text 2. ↵ Antonis Rokas , Barry L Williams , Nicole King , and Sean B Carroll . Genome-scale approaches to resolving incongruence in molecular phylogenies . Nature , 425 ( 6960 ): 798 – 804 , 2003 . OpenUrl CrossRef PubMed Web of Science 3. ↵ Evgenia V Kriventseva , Dmitry Kuznetsov , Fredrik Tegenfeldt , Mosè Manni , Renata Dias , Felipe A Simão , and Evgeny M Zdobnov . OrthoDB v10: sampling the diversity of animal, plant, fungal, protist, bacterial and viral genomes for evolutionary and functional annotations of orthologs . Nucleic Acids Res ., 47 ( D1 ): D807 – D811 , 2019 . OpenUrl CrossRef PubMed 4. ↵ Adrian M Altenhoff , Alex Warwick Vesztrocy , Charles Bernard , Clement-Marie Train , Alina Nicheperovich , Silvia Prieto Baños , Irene Julca , David Moi , Yannis Nevers , Sina Majidian , et al. OMA orthology in 2024: improved prokaryote coverage, ancestral and extant GO enrichment, a revamped synteny viewer and more in the OMA Ecosystem . Nucleic Acids Res ., 52 ( D1 ): D513 – D521 , 2024 . OpenUrl CrossRef PubMed 5. ↵ Jihyeon Kim , Seong-In Na , Dongwook Kim , and Jongsik Chun . UBCG2: Up-to-date bacterial core genes and pipeline for phylogenomic analysis . J. Microbiol ., 59 ( 6 ): 609 – 615 , 2021 . OpenUrl CrossRef PubMed 6. ↵ Dongwook Kim , Cameron LM Gilchrist , Jongsik Chun , and Martin Steinegger . UFCG: database of universal fungal core genes and pipeline for genome-wide phylogenetic analysis of fungi . Nucleic Acids Res ., 51 ( D1 ): D777 – D784 , 2023 . OpenUrl CrossRef PubMed 7. ↵ Adrian M Altenhoff , Natasha M Glover , and Christophe Dessimoz . Inferring Orthology and Paralogy , pages 149 – 175 . Springer New York , New York, NY , 2019 . ISBN 978-1-4939-9074-0 . 8. ↵ L Steven Johnson , Sean R Eddy , and Elon Portugaly . Hidden Markov model speed heuristic and iterative HMM search procedure . BMC Bioinformatics , 11 : 1 – 8 , 2010 . OpenUrl CrossRef PubMed 9. ↵ David M Emms and Steven Kelly . OrthoFinder: phylogenetic orthology inference for comparative genomics . Genome Biol ., 20 : 1 – 14 , 2019 . OpenUrl CrossRef PubMed 10. ↵ Burkhard Rost . Twilight zone of protein sequence alignments . Protein Eng ., 12 ( 2 ): 85 – 94 , 1999 . OpenUrl CrossRef PubMed Web of Science 11. ↵ Nicola Bordin , Christian Dallago , Michael Heinzinger , Stephanie Kim , Maria Littmann , Clemens Rauer , Martin Steinegger , Burkhard Rost , and Christine Orengo . Novel machine learning approaches revolutionize protein knowledge . Trends Biochem. Sci ., 48 ( 4 ): 345 – 359 , 2023 . OpenUrl CrossRef PubMed 12. ↵ Michel Van Kempen , Stephanie S Kim , Charlotte Tumescheit , Milot Mirdita , Jeongjae Lee , Cameron LM Gilchrist , Johannes Söding , and Martin Steinegger . Fast and accurate protein structure search with Fold-seek . Nat. Biotechnol ., 42 ( 2 ): 243 – 246 , 2024 . OpenUrl CrossRef PubMed 13. ↵ S Subbiah , DV Laurents , and M Levitt . Structural similarity of DNA-binding domains of bacteriophage repressors and the globin core . Curr. Biol ., 3 ( 3 ): 141 – 148 , 1993 . OpenUrl CrossRef PubMed Web of Science 14. ↵ Su Yun Chung and S Subbiah . A structural explanation for the twilight zone of protein sequence homology . Structure , 4 ( 10 ): 1123 – 1127 , 1996 . OpenUrl CrossRef PubMed 15. ↵ Inigo Barrio-Hernandez , Jingi Yeo , Jürgen Jänes , Milot Mirdita , Cameron LM Gilchrist , Tanita Wein , Mihaly Varadi , Sameer Velankar , Pedro Beltrao , and Martin Steinegger . Clustering predicted structures at the scale of the known protein universe . Nature , 622 ( 7983 ): 637 – 645 , 2023 . OpenUrl CrossRef PubMed 16. ↵ Martin Steinegger and Johannes Söding . Clustering huge protein sequence sets in linear time . Nat. Commun ., 9 ( 1 ): 2542 , 2018 . OpenUrl CrossRef PubMed 17. ↵ John Jumper , Richard Evans , Alexander Pritzel , Tim Green , Michael Figurnov , Olaf Ronneberger , Kathryn Tunyasuvunakool , Russ Bates , Augustin Žídek , Anna Potapenko , et al. Highly accurate protein structure prediction with AlphaFold . Nature , 596 ( 7873 ): 583 – 589 , 2021 . OpenUrl CrossRef PubMed 18. ↵ Carson M Andorf , Shatabdi Sen , Rita K Hayford , John L Portwood , Ethalinda K Cannon , Lisa C Harper , Jack M Gardiner , Taner Z Sen , and Margaret R Woodhouse . FASSO: An AlphaFold based method to assign functional annotations by combining sequence and structure orthology . bioRxiv , 2022 . 19. ↵ Fabian Ruperti , Nikolaos Papadopoulos , Jacob M Musser , Milot Mirdita , Martin Steinegger , and Detlev Arendt . Cross-phyla protein annotation by structural prediction and alignment . Genome Biol ., 24 ( 1 ): 113 , 2023 . OpenUrl CrossRef PubMed 20. ↵ Caroline Puente-Lelievre , Ashar J Malik , Jordan Douglas , David Ascher , Matthew Baker , Jane Allison , Anthony Poole , Daniel Lundin , Matthew Fullmer , Remco Bouckert , et al. Tertiary-interaction characters enable fast, model-based structural phylogenetics beyond the twilight zone . bioRxiv , 2023 . 21. Jonathon CO Mifsud , Spyros Lytras , Michael R Oliver , Kamilla Toon , Vincenzo A Costa , Edward C Holmes , and Joe Grove . Mapping glycoprotein structure reveals defining events in the evolution of the Flaviviridae . bioRxiv , 2024 . 22. ↵ David Moi , Charles Bernard , Martin Steinegger , Yannis Nevers , Mauricio Langleib , and Christophe Dessimoz . Structural phylogenetics unravels the evolutionary diversification of communication systems in gram-positive bacteria and their viruses . bioRxiv , 2023 . 23. ↵ Joshua Meier , Roshan Rao , Robert Verkuil , Jason Liu , Tom Sercu , and Alex Rives . Language models enable zero-shot prediction of the effects of mutations on protein function . Adv. Neur. In ., 34 : 29287 – 29303 , 2021 . OpenUrl 24. Ahmed Elnaggar , Michael Heinzinger , Christian Dallago , Ghalia Rehawi , Yu Wang , Llion Jones , Tom Gibbs , Tamas Feher , Christoph Angerer , Martin Steinegger , et al. Prottrans: Toward understanding the language of life through self-supervised learning . IEEE T. Pattern Anal ., 44 ( 10 ): 7112 – 7127 , 2021 . OpenUrl 25. ↵ Zeming Lin , Halil Akin , Roshan Rao , Brian Hie , Zhongkai Zhu , Wenting Lu , Nikita Smetanin , Robert Verkuil , Ori Kabeli , Yaniv Shmueli , et al. Evolutionary-scale prediction of atomic-level protein structure with a language model . Science , 379 ( 6637 ): 1123 – 1130 , 2023 . OpenUrl CrossRef PubMed 26. ↵ Michael Heinzinger , Konstantin Weissenow , Joaquin Gomez Sanchez , Adrian Henkel , Milot Mirdita , Martin Steinegger , and Burkhard Rost . Bilingual language model for protein sequence and structure . bioRxiv , 2023 . 27. ↵ Mihaly Varadi , Damian Bertoni , Paulyna Magana , Urmila Paramval , Ivanna Pidruchna , Malarvizhi Radhakrishnan , Maxim Tsenkov , Sreenath Nair , Milot Mirdita , Jingi Yeo , et al. AlphaFold Protein Structure Database in 2024: providing structure coverage for over 214 million protein sequences . Nucleic Acids Res ., 52 ( D1 ): D368 – D375 , 2024 . OpenUrl CrossRef PubMed 28. ↵ Milot Mirdita , Konstantin Schütze , Yoshitaka Moriwaki , Lim Heo , Sergey Ovchinnikov , and Martin Steinegger . ColabFold: making protein folding accessible to all . Nat. Methods , 19 ( 6 ): 679 – 682 , 2022 . OpenUrl CrossRef PubMed 29. ↵ Cameron LM Gilchrist , Milot Mirdita , and Martin Steinegger . Multiple Protein Structure Alignment at Scale with FoldMason . bioRxiv , 2024 . 30. ↵ Bui Quang Minh , Heiko A Schmidt , Olga Chernomor , Dominik Schrempf , Michael D Woodhams , Arndt Von Haeseler , and Robert Lanfear . IQ-TREE 2: new models and efficient methods for phylogenetic inference in the genomic era . Mol. Biol. Evol ., 37 ( 5 ): 1530 – 1534 , 2020 . OpenUrl CrossRef PubMed 31. ↵ Morgan N Price , Paramvir S Dehal , and Adam P Arkin . FastTree 2– approximately maximum-likelihood trees for large alignments . PLoS One , 5 ( 3 ): e9490 , 2010 . OpenUrl CrossRef PubMed 32. ↵ Alexandros Stamatakis . RAxML version 8: a tool for phylogenetic analysis and post-analysis of large phylogenies . Bioinformatics , 30 ( 9 ): 1312 – 1313 , 2014 . OpenUrl CrossRef PubMed Web of Science 33. ↵ UniProt: the universal protein knowledgebase in 2023 . Nucleic Acids Res ., 51 ( D1 ): D523 – D531 , 2023 . OpenUrl CrossRef PubMed 34. ↵ Conrad L Schoch , Stacy Ciufo , Mikhail Domrachev , Carol L Hotton , Sivakumar Kannan , Rogneda Khovanskaya , Detlef Leipe , Richard Mcveigh , Kathleen O’Neill , Barbara Robbertse , et al. NCBI Taxonomy: a comprehensive update on curation, resources and tools . Database , 2020 : baaa062 , 2020 . OpenUrl CrossRef PubMed 35. ↵ Martin Steinegger , Markus Meier , Milot Mirdita , Harald Vöhringer , Stephan J Haunsberger , and Johannes Söding . HH-suite3 for fast remote homology detection and deep protein annotation . BMC Bioinformatics , 20 : 1 – 15 , 2019 . OpenUrl CrossRef PubMed 36. ↵ Kazutaka Katoh and Daron M Standley . MAFFT multiple sequence alignment software version 7: improvements in performance and usability . Mol. Biol. Evol ., 30 ( 4 ): 772 – 780 , 2013 . OpenUrl CrossRef PubMed Web of Science 37. ↵ Fernando Domingues Kümmel Tria , Giddy Landan , and Tal Dagan . Phylogenetic rooting using minimal ancestor deviation . Nat. Ecol. Evol ., 1 ( 7 ): 0193 , 2017 . OpenUrl CrossRef 38. ↵ George F Estabrook , FR McMorris , and Christopher A Meacham . Comparison of undirected phylogenetic trees based on subtrees of four evolutionary units . Syst. Zool ., 34 ( 2 ): 193 – 200 , 1985 . OpenUrl CrossRef 39. ↵ Andreas Sand , Morten K Holt , Jens Johansen , Gerth Stølting Brodal , Thomas Mailund , and Christian NS Pedersen . tqDist: a library for computing the quartet and triplet distances between binary or general trees . Bioinformatics , 30 ( 14 ): 2079 – 2080 , 2014 . OpenUrl CrossRef PubMed 40. ↵ Yoav Benjamini and Yosef Hochberg . Controlling the false discovery rate: a practical and powerful approach to multiple testing . J. R. Stat. Soc. B Met ., 57 ( 1 ): 289 – 300 , 1995 . OpenUrl CrossRef 41. ↵ Mosè Manni , Matthew R Berkeley , Mathieu Seppey , Felipe A Simão , and Evgeny M Zdobnov . BUSCO update: novel and streamlined workflows along with broader and deeper phylogenetic coverage for scoring of eukaryotic, prokaryotic, and viral genomes . Mol. Biol. Evol ., 38 ( 10 ): 4647 – 4654 , 2021 . OpenUrl CrossRef PubMed 42. ↵ Sriram G Garg and Georg KA Hochberg . A general substitution matrix for structural phylogenetics . bioRxiv , 2024 . View the discussion thread. Back to top Previous Next Posted December 22, 2024. Download PDF Supplementary Material Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Unicore Enables Scalable and Accurate Phylogenetic Reconstruction with Structural Core Genes 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 Unicore Enables Scalable and Accurate Phylogenetic Reconstruction with Structural Core Genes Dongwook Kim , Sukhwan Park , Martin Steinegger bioRxiv 2024.12.22.629535; doi: https://doi.org/10.1101/2024.12.22.629535 Share This Article: Copy Citation Tools Unicore Enables Scalable and Accurate Phylogenetic Reconstruction with Structural Core Genes Dongwook Kim , Sukhwan Park , Martin Steinegger bioRxiv 2024.12.22.629535; doi: https://doi.org/10.1101/2024.12.22.629535 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Bioinformatics Subject Areas All Articles Animal Behavior and Cognition (7651) Biochemistry (17746) Bioengineering (13928) Bioinformatics (42066) Biophysics (21499) Cancer Biology (18650) Cell Biology (25579) Clinical Trials (138) Developmental Biology (13409) Ecology (19947) Epidemiology (2067) Evolutionary Biology (24374) Genetics (15633) Genomics (22557) Immunology (17775) Microbiology (40505) Molecular Biology (17217) Neuroscience (88796) Paleontology (667) Pathology (2845) Pharmacology and Toxicology (4836) Physiology (7664) Plant Biology (15179) Scientific Communication and Education (2047) Synthetic Biology (4304) Systems Biology (9839) Zoology (2272)