{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,2,14]],"date-time":"2026-02-14T07:01:11Z","timestamp":1771052471192,"version":"3.50.1"},"reference-count":47,"publisher":"Oxford University Press (OUP)","issue":"2","license":[{"start":{"date-parts":[[2026,1,30]],"date-time":"2026-01-30T00:00:00Z","timestamp":1769731200000},"content-version":"vor","delay-in-days":1,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/501100007316","name":"Klaus Tschira Foundation","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100007316","id-type":"DOI","asserted-by":"publisher"}]},{"name":"the European Union via European Research Council (ERC) Horizon 2020","award":["882500"],"award-info":[{"award-number":["882500"]}]},{"name":"EU ERA Chair","award":["HORIZON-WIDERA-2022-TALENTS-01"],"award-info":[{"award-number":["HORIZON-WIDERA-2022-TALENTS-01"]}]},{"name":"EU ERA Chair","award":["101087081"],"award-info":[{"award-number":["101087081"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,2,3]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:sec>\n                    <jats:title>Motivation<\/jats:title>\n                    <jats:p>Phylogenetic trees describe the evolutionary history among biological species based on their genomic data. Maximum likelihood (ML) based phylogenetic inference tools search for the tree and evolutionary model that best explain the observed genomic data. Given the independence of likelihood score calculations between different genomic sites, parallel computation is commonly deployed. This is followed by a parallel summation over the per-site scores to obtain the overall likelihood score of the tree. However, basic arithmetic operations on IEEE 754 floating-point numbers, such as addition and multiplication, inherently introduce rounding errors. Consequently, the order by which floating-point operations are executed affects the exact resulting likelihood value since these operations are not associative. Moreover, parallel reduction algorithms in numerical codes re-associate operations as a function of the core count and cluster network topology, inducing different round-off errors. These low-level deviations can cause heuristic searches to diverge and induce high-level result discrepancies (e.g. yield topologically distinct phylogenies). This effect has also been observed in multiple scientific fields beyond phylogenetics.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>We observe that varying the degree of parallelism results in diverging phylogenetic tree searches (high-level results) for over 31% out of 10\u00a0179 empirical datasets. More importantly, 8% of these diverging datasets yield trees that are statistically significantly worse than the best-known ML tree for the dataset (AU-test, P\u2009&amp;lt;\u2009.05). To alleviate this, we develop a variant of the widely used phylogenetic inference tool RAxML-NG, which does yield bit-reproducible results under varying core-counts, with a slowdown of only 0%\u201312.7% (median 0.8%) on up to 768 cores. For this, we introduce the ReproRed reduction algorithm, which yields bit-identical results under varying core-counts, by maintaining a fixed operation order that is independent of the communication pattern. ReproRed is thus applicable to all associative reduction operations\u2014in contrast to competitors, which are confined to summation. Our ReproRed reduction algorithm only exchanges the theoretical minimum number of messages, overlaps communication with computation, and utilizes fast base-cases for local reductions. ReproRed is able to all-reduce (via a subsequent broadcast) 4.1\u00d7106 operands across 48\u2013768 cores in 19.7\u201348.61\u2009\u03bcs, thereby exhibiting a slowdown of 13%\u201393% over a non-reproducible all-reduce algorithm. ReproRed outperforms the state-of-the-art reproducible all-reduction algorithm ReproBLAS (offers summation only) beyond 10\u00a0000 elements per core. In summary, we re-assess non-reproducibility in parallel phylogenetic inference, present the first bit-reproducible parallel phylogenetic inference tool, as well as introduce a general algorithm and open-source code for conducting reproducible associative parallel reduction operations.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Availability and implementation<\/jats:title>\n                    <jats:p>ReproRed: https:\/\/doi.org\/10.5281\/zenodo.15004918 (LGPL)\u2014Reproducible RAxML-NG version https:\/\/doi.org\/10.5281\/zenodo.15017407 (GPL)<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1093\/bioinformatics\/btag044","type":"journal-article","created":{"date-parts":[[2026,1,23]],"date-time":"2026-01-23T12:38:49Z","timestamp":1769171929000},"source":"Crossref","is-referenced-by-count":0,"title":["Bit-reproducible parallel phylogenetic tree inference"],"prefix":"10.1093","volume":"42","author":[{"given":"Christoph","family":"Stelz","sequence":"first","affiliation":[{"name":"Karlsruhe Institute of Technology Institute of Theoretical Informatics, , Karlsruhe 76131,","place":["Germany"]},{"name":"Heidelberg Institute for Theoretical Studies Computational Molecular Evolution Group, , Heidelberg 69118,","place":["Germany"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9213-7597","authenticated-orcid":false,"given":"Lukas","family":"H\u00fcbner","sequence":"additional","affiliation":[{"name":"Karlsruhe Institute of Technology Institute of Theoretical Informatics, , Karlsruhe 76131,","place":["Germany"]},{"name":"Heidelberg Institute for Theoretical Studies Computational Molecular Evolution Group, , Heidelberg 69118,","place":["Germany"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Alexandros","family":"Stamatakis","sequence":"additional","affiliation":[{"name":"Foundation for Research and Technology - Hellas Biodiversity Computing Group, Institute of Computer Science, , Heraklion, Crete GR-730013,","place":["Greece"]},{"name":"Karlsruhe Institute of Technology Institute of Theoretical Informatics, , Karlsruhe 76131,","place":["Germany"]},{"name":"Heidelberg Institute for Theoretical Studies Computational Molecular Evolution Group, , Heidelberg 69118,","place":["Germany"]}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"286","published-online":{"date-parts":[[2026,1,29]]},"reference":[{"key":"2026021401013102900_btag044-B1","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3389360","article-title":"Algorithms for efficient reproducible floating point summation","volume":"46","author":"Ahrens","year":"2020","journal-title":"ACM Trans Math Softw"},{"key":"2026021401013102900_btag044-B2","doi-asserted-by":"publisher","first-page":"1235","DOI":"10.1109\/ipdps.2014.127","author":"Arteaga","year":"2014"},{"key":"2026021401013102900_btag044-B3","doi-asserted-by":"publisher","author":"Balaji","year":"2013","DOI":"10.1109\/hpcc.and.euc.2013.65"},{"key":"2026021401013102900_btag044-B4","doi-asserted-by":"publisher","first-page":"71","DOI":"10.1145\/2723872.2723882","article-title":"An introduction to docker for reproducible research","volume":"49","author":"Boettiger","year":"2015","journal-title":"SIGOPS Oper Syst Rev"},{"key":"2026021401013102900_btag044-B5","doi-asserted-by":"publisher","first-page":"e1006650","DOI":"10.1371\/journal.pcbi.1006650","article-title":"Beast 2.5: an advanced software platform for Bayesian evolutionary analysis","volume":"15","author":"Bouckaert","year":"2019","journal-title":"PLoS Comput Biol"},{"key":"2026021401013102900_btag044-B6","doi-asserted-by":"publisher","first-page":"609","DOI":"10.1007\/978-3-319-58943-5_49","author":"Chohra","year":"2017"},{"key":"2026021401013102900_btag044-B7","doi-asserted-by":"publisher","first-page":"83","DOI":"10.1016\/j.parco.2015.09.001","article-title":"Numerical reproducibility for the parallel reduction on multi- and many-core architectures","volume":"49","author":"Collange","year":"2015","journal-title":"Parallel Comput"},{"key":"2026021401013102900_btag044-B8","author":"Corden","year":"2018"},{"key":"2026021401013102900_btag044-B9","doi-asserted-by":"publisher","first-page":"1037","DOI":"10.1093\/molbev\/msy014","article-title":"The state of software for evolutionary biology","volume":"35","author":"Darriba","year":"2018","journal-title":"Mol Biol Evol"},{"key":"2026021401013102900_btag044-B10","first-page":"163","author":"Demmel","year":"2013"},{"key":"2026021401013102900_btag044-B11","doi-asserted-by":"publisher","first-page":"2060","DOI":"10.1109\/tc.2014.2345391","article-title":"Parallel reproducible summation","volume":"64","author":"Demmel","year":"2015","journal-title":"IEEE Trans Comput"},{"key":"2026021401013102900_btag044-B12","doi-asserted-by":"publisher","first-page":"1341","DOI":"10.1007\/s11227-017-2177-5","article-title":"Theoretical peak FLOPS per instruction set: a tutorial","volume":"74","author":"Dolbeau","year":"2018","journal-title":"J Supercomput"},{"key":"2026021401013102900_btag044-B13","doi-asserted-by":"publisher","first-page":"13429","DOI":"10.1073\/pnas.93.23.13429","article-title":"Bootstrap confidence levels for phylogenetic trees","volume":"93","author":"Efron","year":"1996","journal-title":"Proc Natl Acad Sci USA"},{"key":"2026021401013102900_btag044-B14","doi-asserted-by":"publisher","first-page":"368","DOI":"10.1007\/bf01734359","article-title":"Evolutionary trees from DNA sequences: a ml approach","volume":"17","author":"Felsenstein","year":"1981","journal-title":"J Mol Evol"},{"key":"2026021401013102900_btag044-B15","doi-asserted-by":"publisher","first-page":"783","DOI":"10.2307\/2408678","article-title":"Confidence limits on phylogenies: an approach using the bootstrap","volume":"39","author":"Felsenstein","year":"1985","journal-title":"Evolution"},{"key":"2026021401013102900_btag044-B16","doi-asserted-by":"publisher","first-page":"356","DOI":"10.1093\/sysbio\/syu084","article-title":"The phylogenetic likelihood library","volume":"64","author":"Flouri","year":"2015","journal-title":"Syst Biol"},{"key":"2026021401013102900_btag044-B17","doi-asserted-by":"publisher","first-page":"5","DOI":"10.1145\/103162.103163","article-title":"What every computer scientist should know about floating-point arithmetic","volume":"23","author":"Goldberg","year":"1991","journal-title":"ACM Comput Surv"},{"key":"2026021401013102900_btag044-B18","doi-asserted-by":"publisher","DOI":"10.1093\/oxfordjournals.molbev.a040725","article-title":"The number of replications needed for accurate estimation of the bootstrap p value in phylogenetic studies","volume":"9","author":"Hedges","year":"1992","journal-title":"J Mol Evol"},{"key":"2026021401013102900_btag044-B19","doi-asserted-by":"publisher","first-page":"783","DOI":"10.1137\/0914050","article-title":"The accuracy of floating point summation","volume":"14","author":"Higham","year":"1993","journal-title":"SIAM J Sci Comput"},{"key":"2026021401013102900_btag044-B20","doi-asserted-by":"publisher","first-page":"726","DOI":"10.1093\/sysbio\/syw021","article-title":"RevBayes: Bayesian phylogenetic inference using graphical models and an interactive model-specification language","volume":"65","author":"H\u00f6hna","year":"2016","journal-title":"Syst Biol"},{"key":"2026021401013102900_btag044-B21","author":"Iakymchuk","year":"2015"},{"key":"2026021401013102900_btag044-B22","doi-asserted-by":"publisher","author":"IEEE754","year":"2019","DOI":"10.1109\/ieeestd.2019.8766229"},{"key":"2026021401013102900_btag044-B23","author":"Intel","year":"2024"},{"key":"2026021401013102900_btag044-B24","year":"2025"},{"key":"2026021401013102900_btag044-B25","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3186266","article-title":"Reproducibility in scientific computing","volume":"51","author":"Ivie","year":"2018","journal-title":"ACM Comput Surv"},{"key":"2026021401013102900_btag044-B26","doi-asserted-by":"publisher","first-page":"205","DOI":"10.1093\/sysbio\/syw075","article-title":"Efficient detection of repeating sites to accelerate phylogenetic likelihood calculations","volume":"66","author":"Kobert","year":"2017","journal-title":"Syst Biol"},{"key":"2026021401013102900_btag044-B27","doi-asserted-by":"publisher","first-page":"4453","DOI":"10.1093\/bioinformatics\/btz305","article-title":"RAxML-NG: a fast, scalable and user-friendly tool for maximum likelihood phylogenetic inference","volume":"35","author":"Kozlov","year":"2019","journal-title":"Bioinformatics"},{"key":"2026021401013102900_btag044-B28","doi-asserted-by":"publisher","author":"Lei","year":"2023","DOI":"10.1145\/3578178.3578234"},{"key":"2026021401013102900_btag044-B29","doi-asserted-by":"publisher","first-page":"102996","DOI":"10.1016\/j.parco.2023.102996","article-title":"Multi-level parallel multi-layer reproducible summation algorithm","volume":"115","author":"Li","year":"2023","journal-title":"Parallel Comput"},{"key":"2026021401013102900_btag044-B30","article-title":"Docker: lightweight Linux containers for consistent development and deployment","volume":"2014","author":"Merkel","year":"2014","journal-title":"Linux J"},{"key":"2026021401013102900_btag044-B31","doi-asserted-by":"publisher","first-page":"415","DOI":"10.1126\/science.1179653","article-title":"Accessible reproducible research","volume":"327","author":"Mesirov","year":"2010","journal-title":"Science"},{"key":"2026021401013102900_btag044-B33","doi-asserted-by":"publisher","first-page":"81","DOI":"10.1109\/hpcc-smartcity-dss.2017.11","author":"Morel","year":"2017"},{"key":"2026021401013102900_btag044-B34","author":"MPI","year":"2023"},{"key":"2026021401013102900_btag044-B35","author":"Piel","year":"2009"},{"key":"2026021401013102900_btag044-B448024717","doi-asserted-by":"publisher","author":"Priedhorsky","year":":","DOI":"10.1145\/3126908.3126925"},{"key":"2026021401013102900_btag044-B36","doi-asserted-by":"publisher","first-page":"217","DOI":"10.1016\/j.parco.2011.02.009","article-title":"In search of numerical consistency in parallel programming","volume":"37","author":"Robey","year":"2011","journal-title":"Parallel Comput"},{"key":"2026021401013102900_btag044-B37","doi-asserted-by":"publisher","first-page":"131","DOI":"10.1016\/0025-5564(81)90043-2","article-title":"Comparison of phylogenetic trees","volume":"53","author":"Robinson","year":"1981","journal-title":"Math. Biosci"},{"key":"2026021401013102900_btag044-B38","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-25209-0","volume-title":"Sequential and Parallel Algorithms and Data Structures","author":"Sanders","year":"2019"},{"key":"2026021401013102900_btag044-B39","doi-asserted-by":"publisher","first-page":"113","DOI":"10.1111\/j.1096-0031.1989.tb00559.x","article-title":"Confidence limits on phylogenies: the bootstrap revisited","volume":"5","author":"Sanderson","year":"1989","journal-title":"Cladistics"},{"key":"2026021401013102900_btag044-B40","doi-asserted-by":"publisher","first-page":"61","DOI":"10.1109\/5992.881708","article-title":"Making scientific computations reproducible","volume":"2","author":"Schwab","year":"2000","journal-title":"Comput Sci Eng"},{"key":"2026021401013102900_btag044-B41","doi-asserted-by":"publisher","first-page":"6096","DOI":"10.1038\/s41467-020-20005-6","article-title":"An investigation of irreproducibility in maximum likelihood phylogenetic inference","volume":"11","author":"Shen","year":"2020","journal-title":"Nat Commun"},{"key":"2026021401013102900_btag044-B42","doi-asserted-by":"publisher","first-page":"248","DOI":"10.1109\/tdsc.2017.2774299","article-title":"Cross-rack-aware single failure recovery for clustered file systems","volume":"17","author":"Shen","year":"2020","journal-title":"IEEE Trans Dependable Secure Comput"},{"key":"2026021401013102900_btag044-B43","doi-asserted-by":"publisher","first-page":"492","DOI":"10.1080\/10635150290069913","article-title":"An approximately unbiased test of phylogenetic tree selection","volume":"51","author":"Shimodaira","year":"2002","journal-title":"Syst Biol"},{"key":"2026021401013102900_btag044-B44","doi-asserted-by":"publisher","first-page":"639","DOI":"10.3390\/app14020639","article-title":"Bitwise reproducibility for the unstructured computational motif","volume":"14","author":"Sikl\u00f3si","year":"2024","journal-title":"Appl Sci"},{"key":"2026021401013102900_btag044-B45","doi-asserted-by":"publisher","author":"Stamatakis","year":"2003","DOI":"10.1109\/ipdps.2003.1213296"},{"key":"2026021401013102900_btag044-B46","doi-asserted-by":"publisher","first-page":"689","DOI":"10.1109\/sc41406.2024.00050","author":"Uhl","year":"2024"},{"key":"2026021401013102900_btag044-B47","author":"Villa","year":"2009"}],"container-title":["Bioinformatics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/academic.oup.com\/bioinformatics\/advance-article-pdf\/doi\/10.1093\/bioinformatics\/btag044\/66639612\/btag044.pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/bioinformatics\/advance-article-pdf\/doi\/10.1093\/bioinformatics\/btag044\/66639612\/btag044.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/academic.oup.com\/bioinformatics\/advance-article-pdf\/doi\/10.1093\/bioinformatics\/btag044\/66639612\/btag044.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,2,14]],"date-time":"2026-02-14T06:01:40Z","timestamp":1771048900000},"score":1,"resource":{"primary":{"URL":"https:\/\/academic.oup.com\/bioinformatics\/article\/doi\/10.1093\/bioinformatics\/btag044\/8445285"}},"subtitle":[],"editor":[{"given":"Russell","family":"Schwartz","sequence":"additional","affiliation":[],"role":[{"role":"editor","vocabulary":"crossref"}]}],"short-title":[],"issued":{"date-parts":[[2026,1,29]]},"references-count":47,"journal-issue":{"issue":"2","published-print":{"date-parts":[[2026,2,3]]}},"URL":"https:\/\/doi.org\/10.1093\/bioinformatics\/btag044","relation":{},"ISSN":["1367-4803","1367-4811"],"issn-type":[{"value":"1367-4803","type":"print"},{"value":"1367-4811","type":"electronic"}],"subject":[],"published-other":{"date-parts":[[2026,2]]},"published":{"date-parts":[[2026,1,29]]},"article-number":"btag044"}}