{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"institution":[{"name":"bioRxiv"}],"indexed":{"date-parts":[[2026,10,7]],"date-time":"2026-10-07T11:52:05Z","timestamp":1791373925531,"version":"4.3.1"},"posted":{"date-parts":[[2026,8,5]]},"group-title":"Bioinformatics","reference-count":53,"publisher":"openRxiv","license":[{"start":{"date-parts":[[2026,8,5]],"date-time":"2026-08-05T00:00:00Z","timestamp":1785888000000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"accepted":{"date-parts":[[2026,10,2]]},"abstract":"<jats:p>\n                  Cytochrome P450 (CYP) enzymes metabolize roughly three-quarters of clinically used drugs, and genetic variation in these enzymes is a leading source of interindividual differences in drug response. Predicting a variant\u2019s functional effect is therefore critical, yet the consequences of most CYP variants remain unknown. Many state-of-the-art variant-effect predictors rest on a homology-based paradigm that scores variants by evolutionary conservation \u2014 an assumption that pharmacogenes, including CYPs, violate. Focusing on human CYPs, we show that homology-based models fail systematically. AlphaMissense (AM) assigns variants to its \u201cambiguous\u201d class at nearly twice the proteome-wide rate across six clinically important CYPs, and within that class its scores are essentially uninformative about measured CYP2C9 deep mutational scan (DMS) activity (\n                  <jats:italic>\u03c1<\/jats:italic>\n                  = \u22120.130). Using catalytic activity and cellular abundance from the same CYP2C9 DMS, we show that AM is largely an abundance signal: it tracks activity well for destabilizing, low-abundance variants (\n                  <jats:italic>\u03c1<\/jats:italic>\n                  = \u22120.568) but far more weakly for abundance-independent variants (\n                  <jats:italic>\u03c1<\/jats:italic>\n                  = \u22120.357; Fisher\n                  <jats:italic>r-to-z z<\/jats:italic>\n                  = \u22128.99,\n                  <jats:italic>p<\/jats:italic>\n                  &lt; 10\n                  <jats:sup>\u221218<\/jats:sup>\n                  ). We further hypothesized that non-homology-based features (sequence position, substitution chemistry, distance to the substrate-recognition sites, and secondary structure) might account for where AM errs, but their explanatory power is weak. Using CYP2C9 DMS activity as ground truth, we built a\n                  <jats:italic>k<\/jats:italic>\n                  -nearest-neighbors model over ESM-2 embeddings and ensembled it with AM and ESM-2 masked marginal probability, raising the correlation between predicted and measured activity within the ambiguous class from\n                  <jats:italic>\u03c1<\/jats:italic>\n                  = 0.080 for AM alone to 0.399, and reaching an overall\n                  <jats:italic>\u03c1<\/jats:italic>\n                  = 0.721 that exceeds every out-of-the-box predictor we tested (\n                  <jats:italic>\u03c1<\/jats:italic>\n                  = 0.523 to 0.686). However, neither this improved ensemble model nor any other method agrees significantly with clinical annotations. Drawing on evidence that a variant\u2019s effect can depend on the drug, we hypothesize that substrate identity is the key missing feature in current models, and that predicting function for these multi-substrate enzymes may require redefining function as substrate-conditioned.\n                <\/jats:p>","DOI":"10.64898\/2026.07.30.741616","type":"posted-content","created":{"date-parts":[[2026,8,5]],"date-time":"2026-08-05T17:25:23Z","timestamp":1785950723000},"source":"Crossref","is-referenced-by-count":0,"title":["Homology-Based Variant-Effect Predictors Break Down on Cytochrome P450 Pharmacogenes\n                  <sup>*<\/sup>"],"prefix":"10.64898","author":[{"ORCID":"https:\/\/orcid.org\/0009-0001-5766-9130","authenticated-orcid":false,"given":"Helen","family":"Xu","sequence":"first","affiliation":[{"name":"School of Medicine, Stanford University, Stanford, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4859-9863","authenticated-orcid":false,"given":"Issah","family":"Samori","sequence":"additional","affiliation":[{"name":"Department of Bioengineering, Stanford University, Stanford, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5819-7115","authenticated-orcid":false,"given":"Gowri","family":"Nayar","sequence":"additional","affiliation":[{"name":"Department of Biomedical Data Science, Stanford University, Stanford, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-3859-2905","authenticated-orcid":false,"given":"Russ B.","family":"Altman","sequence":"additional","affiliation":[{"name":"School of Medicine, Stanford University, Stanford, CA, USA"},{"name":"Department of Bioengineering, Stanford University, Stanford, CA, USA"},{"name":"Department of Biomedical Data Science, Stanford University, Stanford, CA, USA"},{"name":"Department of Genetics, Stanford University, Stanford, CA, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"54368","reference":[{"key":"2026100704100750000_2026.07.30.741616v2.1","doi-asserted-by":"publisher","DOI":"10.1001\/jama.279.15.1200"},{"key":"2026100704100750000_2026.07.30.741616v2.2","doi-asserted-by":"crossref","unstructured":"J. Sultana , P. Cutroneo and G. Trifir\u00f2 , Clinical and economic burden of adverse drug reactions, Journal of Pharmacology & Pharmacotherapeutics 4 (12 2013).","DOI":"10.4103\/0976-500X.120957"},{"key":"2026100704100750000_2026.07.30.741616v2.3","doi-asserted-by":"crossref","unstructured":"F. P. Guengerich , Human cytochrome P450 enzymes, in Cytochrome P450: Structure, Mechanism, and Biochemistry, Fourth Edition, (Springer International Publishing, 2015) pp. 523\u2013785.","DOI":"10.1007\/978-3-319-12108-6_9"},{"key":"2026100704100750000_2026.07.30.741616v2.4","doi-asserted-by":"crossref","first-page":"103","DOI":"10.1016\/j.pharmthera.2012.12.007","article-title":"Cytochrome P450 enzymes in drug metabolism: Regulation of gene expression, enzyme activities, and impact of genetic variation","volume":"138","year":"2013","journal-title":"Pharmacology and Therapeutics"},{"key":"2026100704100750000_2026.07.30.741616v2.5","doi-asserted-by":"publisher","DOI":"10.1124\/pharmrev.124.001173"},{"key":"2026100704100750000_2026.07.30.741616v2.6","doi-asserted-by":"publisher","DOI":"10.1016\/j.tips.2004.02.007"},{"key":"2026100704100750000_2026.07.30.741616v2.7","doi-asserted-by":"publisher","DOI":"10.7326\/0003-4819-145-10-200611210-00007"},{"key":"2026100704100750000_2026.07.30.741616v2.8","doi-asserted-by":"publisher","DOI":"10.1016\/j.ajhg.2021.07.001"},{"key":"2026100704100750000_2026.07.30.741616v2.9","doi-asserted-by":"publisher","DOI":"10.1038\/s41576-022-00572-8"},{"key":"2026100704100750000_2026.07.30.741616v2.10","doi-asserted-by":"crossref","first-page":"91","DOI":"10.1146\/annurev-pharmtox-061724-080935","article-title":"Pharmacogenetic Panel Testing: A Review of Current Practice and Potential for Clinical Implementation","volume":"65","year":"2025","journal-title":"Annual Review of Pharmacology and Toxicology"},{"key":"2026100704100750000_2026.07.30.741616v2.11","first-page":"1","article-title":"Sift missense predictions for genomes","volume":"11","year":"2016","journal-title":"Nature protocols"},{"key":"2026100704100750000_2026.07.30.741616v2.12","doi-asserted-by":"crossref","unstructured":"J. Cheng , G. Novati , J. Pan , C. Bycroft , A. \u017demgulyte , T. Applebaum , A. Pritzel , L. H. Wong , M. Zielinski , T. Sargeant , R. G. Schneider , A. W. Senior , J. Jumper , D. Hassabis , P. Kohli and \u017d. Avsec , Accurate proteome-wide missense variant effect prediction with AlphaMissense, Science 381 (9 2023).","DOI":"10.1126\/science.adg7492"},{"key":"2026100704100750000_2026.07.30.741616v2.13","doi-asserted-by":"publisher","DOI":"10.1038\/S41586-021-03819-2"},{"key":"2026100704100750000_2026.07.30.741616v2.14","doi-asserted-by":"publisher","DOI":"10.1126\/science.ade2574"},{"key":"2026100704100750000_2026.07.30.741616v2.15","doi-asserted-by":"publisher","DOI":"10.1038\/s41588-023-01465-0"},{"key":"2026100704100750000_2026.07.30.741616v2.16","doi-asserted-by":"crossref","unstructured":"J. Meier , R. Rao , R. Verkuil , J. Liu , T. Sercu and A. Rives , Language models enable zero-shot prediction of the effects of mutations on protein function, in Advances in Neural Information Processing Systems, eds. M. Ranzato , A. Beygelzimer , Y. Dauphin , P. S. Liang and J. W. Vaughan (Curran Associates, Inc., 2021).","DOI":"10.1101\/2021.07.09.450648"},{"key":"2026100704100750000_2026.07.30.741616v2.17","doi-asserted-by":"publisher","DOI":"10.1016\/j.ajhg.2016.08.016"},{"key":"2026100704100750000_2026.07.30.741616v2.18","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btaa1030"},{"key":"2026100704100750000_2026.07.30.741616v2.19","doi-asserted-by":"publisher","DOI":"10.1038\/s41397-026-00399-0"},{"key":"2026100704100750000_2026.07.30.741616v2.20","doi-asserted-by":"crossref","first-page":"evac167","DOI":"10.1093\/gbe\/evac167","article-title":"Joint Analysis of Phenotypic and Genomic Diversity Sheds Light on the Evolution of Xenobiotic Metabolism in Humans","volume":"14","year":"2022","journal-title":"Genome Biology and Evolution"},{"key":"2026100704100750000_2026.07.30.741616v2.21","first-page":"557","article-title":"Evolution of the P450 gene superfamily","volume":"274","year":"1988","journal-title":"Progress in Clinical and Biological Research"},{"key":"2026100704100750000_2026.07.30.741616v2.22","first-page":"115","article-title":"An optimized prediction framework to assess the functional impact of pharmacogenetic variants","volume":"19","year":"2018","journal-title":"The Pharmacogenomics Journal"},{"key":"2026100704100750000_2026.07.30.741616v2.23","doi-asserted-by":"publisher","DOI":"10.1038\/s41397-024-00338-x"},{"key":"2026100704100750000_2026.07.30.741616v2.24","doi-asserted-by":"crossref","first-page":"171","DOI":"10.1080\/14622416.2025.2504863","article-title":"Machine learning models for pharmacogenomic variant effect predictions \u2013recent developments and future frontiers","volume":"26","year":"2025","journal-title":"Pharmacogenomics"},{"key":"2026100704100750000_2026.07.30.741616v2.25","doi-asserted-by":"crossref","first-page":"e1008399","DOI":"10.1371\/journal.pcbi.1008399","article-title":"Transfer learning enables prediction of CYP2D6 haplotype function","volume":"16","year":"2020","journal-title":"PLOS Computational Biology"},{"key":"2026100704100750000_2026.07.30.741616v2.26","doi-asserted-by":"publisher","DOI":"10.1093\/genetics\/iyae156"},{"key":"2026100704100750000_2026.07.30.741616v2.27","first-page":"D980","article-title":"ClinVar: public archive of relationships among sequence variation and human phenotype","volume":"42","year":"2013","journal-title":"Nucleic Acids Research"},{"key":"2026100704100750000_2026.07.30.741616v2.28","doi-asserted-by":"publisher","DOI":"10.1038\/gim.2015.30"},{"key":"2026100704100750000_2026.07.30.741616v2.29","doi-asserted-by":"crossref","first-page":"542","DOI":"10.1002\/cpt.2321","article-title":"PharmVar: A Global Resource and Repository for Pharmacogene Variation","volume":"110","year":"2021","journal-title":"Clinical Pharmacology & Therapeutics"},{"key":"2026100704100750000_2026.07.30.741616v2.30","doi-asserted-by":"crossref","first-page":"563","DOI":"10.1002\/cpt.2350","article-title":"An Evidence-Based Framework for Evaluating Pharmacogenomics Knowledge for Personalized Medicine","volume":"110","year":"2021","journal-title":"Clinical Pharmacology & Therapeutics"},{"key":"2026100704100750000_2026.07.30.741616v2.31","doi-asserted-by":"publisher","DOI":"10.1093\/nar\/gkaa1100"},{"key":"2026100704100750000_2026.07.30.741616v2.32","doi-asserted-by":"crossref","unstructured":"X. Liu , C. Li , C. Mou , Y. Dong and Y. Tu , dbnsfp v4: a comprehensive database of transcript-specific functional predictions and annotations for human nonsynonymous and splice-site snvs, Genome medicine 12 (12 2020).","DOI":"10.1186\/s13073-020-00803-9"},{"key":"2026100704100750000_2026.07.30.741616v2.33","doi-asserted-by":"crossref","first-page":"1593","DOI":"10.1002\/prot.24840","article-title":"Identification of universal selectivity-determining positions in cytochrome p450 monooxygenases by systematic sequence-based literature mining","volume":"83","year":"2015","journal-title":"Proteins: Structure, Function and Bioinformatics"},{"key":"2026100704100750000_2026.07.30.741616v2.34","doi-asserted-by":"crossref","first-page":"12345","DOI":"10.1038\/s41598-021-91494-8","article-title":"Uncovering of cytochrome P450 anatomy by SecStrAnnotator","volume":"11","year":"2021","journal-title":"Scientific Reports"},{"key":"2026100704100750000_2026.07.30.741616v2.35","doi-asserted-by":"publisher","DOI":"10.1016\/j.ajhg.2025.01.012"},{"key":"2026100704100750000_2026.07.30.741616v2.36","first-page":"111071","article-title":"Varland: A pipeline to map the structural landscape of missense variants at the proteome scale","volume":"302","year":"2025","journal-title":"Journal of Biological Chemistry"},{"key":"2026100704100750000_2026.07.30.741616v2.37","doi-asserted-by":"crossref","first-page":"7189","DOI":"10.1038\/s41467-025-62318-4","article-title":"Premode predicts mode-of-action of missense variants by deep graph representation learning of protein sequence and structural context","volume":"16","year":"2025","journal-title":"Nature Communications"},{"key":"2026100704100750000_2026.07.30.741616v2.38","doi-asserted-by":"crossref","first-page":"501","DOI":"10.1038\/s41397-019-0105-1","article-title":"Challenges to assess substrate-dependent allelic effects in CYP450 enzymes and the potential clinical implications","volume":"19","year":"2019","journal-title":"The Pharmacogenomics Journal"},{"key":"2026100704100750000_2026.07.30.741616v2.39","doi-asserted-by":"crossref","first-page":"3673","DOI":"10.3390\/molecules30183673","article-title":"The Activation of Cytochrome P450 2C9 Is Facilitated by the Coenzyme Forms of Vitamin B2","volume":"30","year":"2025","journal-title":"Molecules"},{"key":"2026100704100750000_2026.07.30.741616v2.40","unstructured":"A. Lafita , F. Gonzalez , M. Hossam , P. Smyth , J. Deasy , A. Allyn-Feuer , D. Seaton and S. Young , Fine-tuning Protein Language Models with Deep Mutational Scanning improves Variant Effect Prediction, arXiv preprint arXiv:2405.06729 (5 2024)."},{"key":"2026100704100750000_2026.07.30.741616v2.41","doi-asserted-by":"crossref","unstructured":"M. Glaser and J. Br\u00e4gelmann , ESM-Effect: An Effective and Efficient Fine-Tuning Framework towards accurate prediction of Mutation\u2019s Functional Effect, bioRxiv, p. 2025.02.03.635741 (2 2025).","DOI":"10.1101\/2025.02.03.635741"},{"key":"2026100704100750000_2026.07.30.741616v2.42","doi-asserted-by":"crossref","unstructured":"T. Deguchi , N. S. A. Ghani , Y. Kurumida , S. Iida , K. Kobayashi and Y. Saito , Data-efficient protein mutational effect prediction with weak supervision by molecular simulation and protein language models, Briefings in Bioinformatics 26 (8 2025).","DOI":"10.1093\/bib\/bbaf536"},{"key":"2026100704100750000_2026.07.30.741616v2.43","doi-asserted-by":"crossref","unstructured":"H. Kim , E. Lee , S. Chung , J. Park , S. Baek and J. Kang , ATTNSOM: Learning Cross-Isoform Attention for Cytochrome P450 Site-of-Metabolism, arXiv preprint arXiv:2601.20891 (1 2026).","DOI":"10.1093\/bioinformatics\/btag314"},{"key":"2026100704100750000_2026.07.30.741616v2.44","doi-asserted-by":"crossref","first-page":"3149","DOI":"10.1021\/acs.jcim.4c00115","article-title":"DeepP450: Predicting Human P450 Activities of Small Molecules by Integrating Pretrained Protein Language Model and Molecular Representation","volume":"64","year":"2024","journal-title":"Journal of Chemical Information and Modeling"},{"key":"2026100704100750000_2026.07.30.741616v2.45","doi-asserted-by":"publisher","DOI":"10.1021\/acs.jcim.4c02051"},{"key":"2026100704100750000_2026.07.30.741616v2.46","doi-asserted-by":"crossref","unstructured":"Z. Nie , H. Zhang , H. Jiang , Y. Liu , X. Huang , F. Xu , J. Fu , Z. Ren , Y. Tian , W.-B. Zhang and J. Chen , OmniESI: A unified framework for enzyme-substrate interaction prediction with progressive conditional deep learning, arXiv preprint arXiv:2506.17963 (2025).","DOI":"10.21203\/rs.3.rs-5516445\/v1"},{"key":"2026100704100750000_2026.07.30.741616v2.47","doi-asserted-by":"publisher","DOI":"10.1038\/s41929-022-00798-z"},{"key":"2026100704100750000_2026.07.30.741616v2.48","doi-asserted-by":"crossref","first-page":"8211","DOI":"10.1038\/s41467-023-44113-1","article-title":"UniKP: a unified framework for the prediction of enzyme kinetic parameters","volume":"14","year":"2023","journal-title":"Nature Communications"},{"key":"2026100704100750000_2026.07.30.741616v2.49","doi-asserted-by":"crossref","first-page":"218","DOI":"10.1002\/cpt.3680","article-title":"Improved Prediction of CYP2D6 Catalyzed Drug Metabolism by Taking Variant Substrate Specificities and Novel Polymorphic Haplotypes into Account","volume":"118","year":"2025","journal-title":"Clinical Pharmacology and Therapeutics"},{"key":"2026100704100750000_2026.07.30.741616v2.50","doi-asserted-by":"crossref","first-page":"e70592","DOI":"10.1111\/cts.70592","article-title":"Feed-Forward Deep Neural Networks Predict Substrate-Specific Effects of Transporter Variants to Explain Drug Response Variability","volume":"19","year":"2026","journal-title":"Clinical and Translational Science"},{"key":"2026100704100750000_2026.07.30.741616v2.51","doi-asserted-by":"publisher","DOI":"10.1021\/bi801231m"},{"key":"2026100704100750000_2026.07.30.741616v2.52","doi-asserted-by":"publisher","DOI":"10.1124\/jpet.109.150706"},{"key":"2026100704100750000_2026.07.30.741616v2.53","doi-asserted-by":"crossref","first-page":"3245","DOI":"10.1038\/s41467-026-69913-z","article-title":"Decoding the substrate specificity landscape of a promiscuous enzyme through multi-substrate mutational scanning","volume":"17","year":"2026","journal-title":"Nature Communications"}],"container-title":[],"original-title":[],"link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.64898\/2026.07.30.741616","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,10,7]],"date-time":"2026-10-07T11:10:21Z","timestamp":1791371421000},"score":1,"resource":{"primary":{"URL":"http:\/\/biorxiv.org\/lookup\/doi\/10.64898\/2026.07.30.741616"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,8,5]]},"references-count":53,"URL":"https:\/\/doi.org\/10.64898\/2026.07.30.741616","relation":{},"subject":[],"published":{"date-parts":[[2026,8,5]]},"subtype":"preprint"}}