{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"institution":[{"name":"bioRxiv"}],"indexed":{"date-parts":[[2026,9,12]],"date-time":"2026-09-12T19:05:06Z","timestamp":1789239906775,"version":"build-2803163510"},"posted":{"date-parts":[[2026,9,10]]},"group-title":"Bioinformatics","reference-count":62,"publisher":"openRxiv","license":[{"start":{"date-parts":[[2026,9,10]],"date-time":"2026-09-10T00:00:00Z","timestamp":1788998400000},"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":[[2026,9,10]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                <jats:p>Combination therapies are central to cancer treatment, but exhaustive screening is impractical. We introduce PHAROS, a framework that turns a pretrained single-cell perturbation model into a target-directed search engine for drug combinations. PHAROS predicts how a cell population changes under a drug, one drug at a time, then chains these predictions together to simulate drug combinations. It scores each simulated outcome against the desired target state and uses a search algorithm to find the most promising combinations, all without retraining the underlying model. Across two independent combinatorial perturbation datasets, PHAROS recovered exact or mechanism-matched two-drug responses in cell lines, both seen and unseen during model training. Its rankings were specific to the requested conversion and were not explained by single-drug effects, additive effects, or shared mechanism of action. In exploratory analyses of patient-derived metastatic HR+\/HER2 breast tumors and basal cell carcinoma (BCC), PHAROS prioritized FDA-approved regimens, distinguished combinations by their predicted tumor-versus-immune objective profiles, and nominated pathway-level hypotheses, while explicitly identifying both tumor cohorts as outside the model\u2019s supported distribution. PHAROS provides a modular route from pretrained virtual-cell models to inverse, single-cell combination-screening platforms.<\/jats:p>","DOI":"10.64898\/2026.09.08.749477","type":"posted-content","created":{"date-parts":[[2026,9,11]],"date-time":"2026-09-11T01:45:14Z","timestamp":1789091114000},"source":"Crossref","is-referenced-by-count":0,"title":["PHAROS: turning single-cell perturbation models into target-directed drug-combination screens"],"prefix":"10.64898","author":[{"given":"Jon","family":"Bezney","sequence":"first","affiliation":[{"name":"Department of Genetics, Stanford University School of Medicine, Stanford, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Carlo","family":"Ruggeri","sequence":"additional","affiliation":[{"name":"Department of Computing Sciences, Bocconi Institute for Data Science and Analytics, Milan, Italy"},{"name":"Department of Statistics, University of British Columbia, Vancouver, Canada"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Federico","family":"Borra","sequence":"additional","affiliation":[{"name":"Department of Computing Sciences, Bocconi Institute for Data Science and Analytics, Milan, Italy"},{"name":"IFOM ETS \u2013 The AIRC Institute of Molecular Oncology, Milan, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lei S.","family":"Qi","sequence":"additional","affiliation":[{"name":"Department of Bioengineering, Stanford University, Stanford, USA"},{"name":"Sarafan ChEM-H, Stanford University, Stanford, USA"},{"name":"Chan Zuckerberg Biohub-San Francisco, San Francisco, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-0409-406X","authenticated-orcid":false,"given":"Francesca","family":"Buffa","sequence":"additional","affiliation":[{"name":"Department of Computing Sciences, Bocconi Institute for Data Science and Analytics, Milan, Italy"},{"name":"IFOM ETS \u2013 The AIRC Institute of Molecular Oncology, Milan, Italy"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lars M.","family":"Steinmetz","sequence":"additional","affiliation":[{"name":"Department of Genetics, Stanford University School of Medicine, Stanford, USA"},{"name":"Genome Biology Unit, European Molecular Biology Laboratory (EMBL), Heidelberg, Germany"},{"name":"Stanford Genome Technology Center, Palo Alto, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"54368","reference":[{"key":"2026091211100999000_2026.09.08.749477v1.1","first-page":"715","article-title":"scgen predicts single-cell perturbation responses","volume":"16","year":"2019","journal-title":"Nature Methods 2019 16:8"},{"key":"2026091211100999000_2026.09.08.749477v1.2","first-page":"1759","article-title":"Learning single-cell perturbation responses using neural optimal transport","volume":"20","year":"2023","journal-title":"Nature Methods 2023 20:11"},{"key":"2026091211100999000_2026.09.08.749477v1.3","first-page":"1769","article-title":"Causal identification of single-cell experimental perturbation effects with cinema-ot","volume":"20","year":"2023","journal-title":"Nature Methods 2023 20:11"},{"key":"2026091211100999000_2026.09.08.749477v1.4","doi-asserted-by":"publisher","DOI":"10.1093\/bioinformatics\/btaf253"},{"key":"2026091211100999000_2026.09.08.749477v1.5","doi-asserted-by":"crossref","first-page":"MSB202211517","DOI":"10.15252\/msb.202211517","article-title":"Predicting cellular responses to complex perturbations in high-throughput screens","volume":"19","year":"2023","journal-title":"Molecular Systems Biology"},{"key":"2026091211100999000_2026.09.08.749477v1.6","doi-asserted-by":"publisher","DOI":"10.1038\/s44320-025-00131-3"},{"key":"2026091211100999000_2026.09.08.749477v1.7","first-page":"1470","article-title":"scgpt: toward building a foundation model for single-cell multi-omics using generative ai","volume":"21","year":"2024","journal-title":"Nature Methods 2024 21:8"},{"key":"2026091211100999000_2026.09.08.749477v1.8","unstructured":"Adduri, A. K. et al. Predicting cellular responses to perturbation across diverse contexts with state. bioRxiv 2025.06.26.661135 (2025). URL https:\/\/www.biorxiv.org\/content\/10.1101\/2025.06.26.661135v2."},{"key":"2026091211100999000_2026.09.08.749477v1.9","unstructured":"Dong, M. et al. Stack: In-context learning of single-cell biology. bioRxiv 2026.01.09.698608 (2026). URL https:\/\/www.biorxiv.org\/content\/10.64898\/2026.01.09.698608v2."},{"key":"2026091211100999000_2026.09.08.749477v1.10","first-page":"984","article-title":"Conditional monge gap enables generalizable single-cell perturbation modelling","volume":"8","year":"2026","journal-title":"Nature Machine Intelligence 2026 8:6"},{"key":"2026091211100999000_2026.09.08.749477v1.11","first-page":"451","article-title":"Benchmarking algorithms for generalizable single-cell perturbation response prediction","volume":"23","year":"2025","journal-title":"Nature Methods 2025 23:2"},{"key":"2026091211100999000_2026.09.08.749477v1.12","first-page":"1008","article-title":"Combinatorial prediction of therapeutic perturbations using causally inspired neural networks","volume":"10","year":"2025","journal-title":"Nature Biomedical Engineering 2025 10:5"},{"key":"2026091211100999000_2026.09.08.749477v1.13","first-page":"16015","article-title":"L1000cds2: Lincs l1000 characteristic direction signatures search engine","volume":"2","year":"2016","journal-title":"npj Systems Biology and Applications 2016 2:1"},{"key":"2026091211100999000_2026.09.08.749477v1.14","doi-asserted-by":"crossref","unstructured":"Zhang, J. et al. Tahoe-100m: A giga-scale single-cell perturbation atlas for context-dependent gene function and cellular modeling. bioRxiv 2025.02.20.639398 (2025). URL https:\/\/www.biorxiv.org\/content\/10.1101\/2025.02.20.639398v3.","DOI":"10.1101\/2025.02.20.639398"},{"key":"2026091211100999000_2026.09.08.749477v1.15","doi-asserted-by":"publisher","DOI":"10.1016\/S0169-7439(01)00155-1"},{"key":"2026091211100999000_2026.09.08.749477v1.16","unstructured":"Cuturi, M. Sinkhorn distances: Lightspeed computation of optimal transport. Advances in Neural Information Processing Systems 26 (2013)."},{"key":"2026091211100999000_2026.09.08.749477v1.17","doi-asserted-by":"crossref","unstructured":"Vijayakumar, A. K. et al. Diverse beam search for improved description of complex scenes. Proceedings of the AAAI Conference on Artificial Intelligence 32, 7371\u20137379 (2018). URL https:\/\/ojs.aaai.org\/index.php\/AAAI\/article\/view\/12340.","DOI":"10.1609\/aaai.v32i1.12340"},{"key":"2026091211100999000_2026.09.08.749477v1.18","doi-asserted-by":"crossref","first-page":"100487","DOI":"10.1016\/j.xgen.2023.100487","article-title":"Multiplex single-cell chemical genomics reveals the kinase dependence of the response to targeted therapy","volume":"4","year":"2024","journal-title":"Cell Genomics"},{"key":"2026091211100999000_2026.09.08.749477v1.19","first-page":"95","article-title":"Single-cell rna sequencing reveals different cellular states in malignant cells and the tumor microenvironment in primary and metastatic er-positive breast cancer","volume":"11","year":"2025","journal-title":"npj Breast Cancer 2025 11:1"},{"key":"2026091211100999000_2026.09.08.749477v1.20","doi-asserted-by":"crossref","first-page":"1023","DOI":"10.1016\/S0140-6736(17)33326-3","article-title":"Global surveillance of trends in cancer survival 2000\u201314 (concord-3): analysis of individual records for 37513025 patients diagnosed with one of 18 cancers from 322 population-based registries in 71 countries","volume":"391","year":"2018","journal-title":"The Lancet"},{"key":"2026091211100999000_2026.09.08.749477v1.21","doi-asserted-by":"publisher","DOI":"10.1016\/j.cell.2011.09.024"},{"key":"2026091211100999000_2026.09.08.749477v1.22","doi-asserted-by":"crossref","first-page":"188595","DOI":"10.1016\/j.bbcan.2021.188595","article-title":"Fgfr signalling and endocrine resistance in breast cancer: Challenges for the clinical development of fgfr inhibitors","volume":"1876","year":"2021","journal-title":"Biochimica et biophysica acta. Reviews on cancer"},{"key":"2026091211100999000_2026.09.08.749477v1.23","doi-asserted-by":"crossref","first-page":"2775","DOI":"10.3390\/biomedicines10112775","article-title":"Exploiting dna replication stress as a therapeutic strategy for breast cancer","volume":"10","year":"2022","journal-title":"Biomedicines"},{"key":"2026091211100999000_2026.09.08.749477v1.24","first-page":"16042","article-title":"The paradoxical functions of egfr during breast cancer progression","volume":"2","year":"2017","journal-title":"Signal Transduction and Targeted Therapy 2017 2:1"},{"key":"2026091211100999000_2026.09.08.749477v1.25","doi-asserted-by":"publisher","DOI":"10.1038\/ejhg.2009.230"},{"key":"2026091211100999000_2026.09.08.749477v1.26","doi-asserted-by":"crossref","first-page":"3029","DOI":"10.3390\/cancers12103029","article-title":"The fgf\/fgfr system in breast cancer: Oncogenic features and therapeutic perspectives","volume":"12","year":"2020","journal-title":"Cancers"},{"key":"2026091211100999000_2026.09.08.749477v1.27","doi-asserted-by":"crossref","first-page":"159","DOI":"10.1007\/978-981-32-9620-6_8","article-title":"Dna damage repair inhibitor for breast cancer treatment","volume":"1187","year":"2021","journal-title":"Advances in experimental medicine and biology"},{"key":"2026091211100999000_2026.09.08.749477v1.28","first-page":"245","article-title":"Egfr and her2 signaling in breast cancer brain metastasis","volume":"8","year":"2016","journal-title":"Frontiers in bioscience (Elite edition)"},{"key":"2026091211100999000_2026.09.08.749477v1.29","doi-asserted-by":"crossref","first-page":"115877","DOI":"10.1016\/j.biopha.2023.115877","article-title":"Targeting the dna repair pathway for breast cancer therapy: Beyond the molecular subtypes","volume":"169","year":"2023","journal-title":"Biomedicine & Pharmacotherapy"},{"key":"2026091211100999000_2026.09.08.749477v1.30","doi-asserted-by":"crossref","first-page":"111","DOI":"10.1186\/s40364-024-00653-2","article-title":"Dna damage response in breast cancer and its significant role in guiding novel precise therapies","volume":"12","year":"2024","journal-title":"Biomarker Research"},{"key":"2026091211100999000_2026.09.08.749477v1.31","doi-asserted-by":"publisher","DOI":"10.1038\/s41591-019-0522-3"},{"key":"2026091211100999000_2026.09.08.749477v1.32","doi-asserted-by":"publisher","DOI":"10.1038\/s41577-019-0218-4"},{"key":"2026091211100999000_2026.09.08.749477v1.33","doi-asserted-by":"publisher","DOI":"10.1042\/EBC20230001"},{"key":"2026091211100999000_2026.09.08.749477v1.34","doi-asserted-by":"publisher","DOI":"10.1056\/NEJMoa1604958"},{"key":"2026091211100999000_2026.09.08.749477v1.35","first-page":"43","article-title":"Cancer cells resistant to immune checkpoint blockade acquire interferon-associated epigenetic memory to sustain t cell dysfunction","volume":"4","year":"2023","journal-title":"Nature Cancer 2022 4:1"},{"key":"2026091211100999000_2026.09.08.749477v1.36","doi-asserted-by":"publisher","DOI":"10.1016\/j.cell.2016.11.022"},{"key":"2026091211100999000_2026.09.08.749477v1.37","doi-asserted-by":"publisher","DOI":"10.1038\/s41577-018-0029-z"},{"key":"2026091211100999000_2026.09.08.749477v1.38","doi-asserted-by":"crossref","unstructured":"Luo, N. , et al. Melanoma response to anti-pd-l1 immunotherapy requires jak1 signaling, but not jak2. OncoImmunology 7 (2018). URL https:\/\/www.tandfonline.com\/doi\/pdf\/10.1080\/2162402X.2018.1438106.","DOI":"10.1080\/2162402X.2018.1438106"},{"key":"2026091211100999000_2026.09.08.749477v1.39","doi-asserted-by":"crossref","first-page":"1790936","DOI":"10.3389\/fimmu.2026.1790936","article-title":"Jak inhibition in pd-1 immunotherapy and tumor microenvironment","volume":"17","year":"2026","journal-title":"Frontiers in Immunology"},{"key":"2026091211100999000_2026.09.08.749477v1.40","doi-asserted-by":"crossref","unstructured":"Arias-Badia, M. , et al. Sequential jak inhibition enhances antitumor immunity after combined anti\u2013pd-1 and anti-ctla4. JCI Insight 10 (2025). URL https:\/\/insight.jci.org\/articles\/view\/187921.","DOI":"10.1172\/jci.insight.187921"},{"key":"2026091211100999000_2026.09.08.749477v1.41","doi-asserted-by":"crossref","first-page":"1602752","DOI":"10.3389\/fimmu.2025.1602752","article-title":"Plk1 in cancer therapy: a comprehensive review of immunomodulatory mechanisms and therapeutic opportunities","volume":"16","year":"2025","journal-title":"Frontiers in Immunology"},{"key":"2026091211100999000_2026.09.08.749477v1.42","doi-asserted-by":"publisher","DOI":"10.3389\/fonc.2022.903016\/full"},{"key":"2026091211100999000_2026.09.08.749477v1.43","first-page":"3979527","article-title":"Exploration of the combination of plk1 inhibition with immunotherapy in cancer treatment","volume":"2018","year":"2018","journal-title":"Journal of Oncology"},{"key":"2026091211100999000_2026.09.08.749477v1.44","doi-asserted-by":"crossref","unstructured":"Su, S. , Chhabra, G. , Singh, C. K. , Ndiaye, M. A. & Ahmad, N. Plk1 inhibition-based combination therapies for cancer management. Translational Oncology 16 (2022). URL https:\/\/pubmed.ncbi.nlm.nih.gov\/34973570\/.","DOI":"10.1016\/j.tranon.2021.101332"},{"key":"2026091211100999000_2026.09.08.749477v1.45","doi-asserted-by":"crossref","first-page":"3516","DOI":"10.7150\/ijms.60135","article-title":"Plk1 inhibition induces immunogenic cell death and enhances immunity against nsclc","volume":"18","year":"2021","journal-title":"International Journal of Medical Sciences"},{"key":"2026091211100999000_2026.09.08.749477v1.46","doi-asserted-by":"publisher","DOI":"10.1158\/0008-5472.CAN-22-0018"},{"key":"2026091211100999000_2026.09.08.749477v1.47","first-page":"4261","article-title":"Development of a nanoparticle-based immunotherapy targeting pd-l1 and plk1 for lung cancer treatment","volume":"13","year":"2022","journal-title":"Nature Communications 2022 13:1"},{"key":"2026091211100999000_2026.09.08.749477v1.48","doi-asserted-by":"publisher","DOI":"10.1146\/annurev-immunol-070621-030155"},{"key":"2026091211100999000_2026.09.08.749477v1.49","doi-asserted-by":"publisher","DOI":"10.1097\/PPO.0000000000000303"},{"key":"2026091211100999000_2026.09.08.749477v1.50","doi-asserted-by":"crossref","first-page":"798474","DOI":"10.3389\/fimmu.2021.798474","article-title":"Novel molecular determinants of response or resistance to immune checkpoint inhibitor therapies in melanoma","volume":"12","year":"2022","journal-title":"Frontiers in Immunology"},{"key":"2026091211100999000_2026.09.08.749477v1.51","doi-asserted-by":"crossref","first-page":"60975","DOI":"10.18632\/oncotarget.17689","article-title":"Genetic defects of the irf1-mediated major histocompatibility complex class i antigen presentation pathway occur prevalently in the jak2 gene in non-small cell lung cancer","volume":"8","year":"2017","journal-title":"Oncotarget"},{"key":"2026091211100999000_2026.09.08.749477v1.52","doi-asserted-by":"crossref","first-page":"1688699","DOI":"10.3389\/fimmu.2025.1688699","article-title":"Overcoming resistance to pd-1 and ctla-4 blockade mechanisms and therapeutic strategies","volume":"16","year":"2025","journal-title":"Frontiers in Immunology"},{"key":"2026091211100999000_2026.09.08.749477v1.53","first-page":"A557","article-title":"Identifying genetic factors for resistance and response to immune checkpoint inhibitors (icis) in advanced non-small cell lung cancer (nsclc) patients using real-world data","volume":"10","year":"2022","journal-title":"Journal for ImmunoTherapy of Cancer"},{"key":"2026091211100999000_2026.09.08.749477v1.54","doi-asserted-by":"publisher","DOI":"10.1126\/science.aax6234"},{"key":"2026091211100999000_2026.09.08.749477v1.55","first-page":"531","article-title":"scperturb: harmonized single-cell perturbation data","volume":"21","year":"2024","journal-title":"Nature Methods 2024 21:3"},{"key":"2026091211100999000_2026.09.08.749477v1.56","first-page":"1088","article-title":"The curated cancer cell atlas provides a comprehensive characterization of tumors at single-cell resolution","volume":"6","year":"2025","journal-title":"Nature Cancer 2025 6:6"},{"key":"2026091211100999000_2026.09.08.749477v1.57","doi-asserted-by":"crossref","unstructured":"Reveron-Thornton, R. F. et al. A pan-cancer single-cell atlas to evaluate tumor identity, cell line concordance, and dependency mapping. bioRxiv 2026.02.14.705396 (2026). URL https:\/\/www.biorxiv.org\/content\/10.64898\/2026.02.14.705396v2.","DOI":"10.64898\/2026.02.14.705396"},{"key":"2026091211100999000_2026.09.08.749477v1.58","doi-asserted-by":"publisher","DOI":"10.1038\/s41586-023-06139-9"},{"key":"2026091211100999000_2026.09.08.749477v1.59","first-page":"1481","article-title":"Large-scale foundation model on single-cell transcriptomics","volume":"21","year":"2024","journal-title":"Nature Methods 2024 21:8"},{"key":"2026091211100999000_2026.09.08.749477v1.60","unstructured":"Rosen, Y. et al. Universal cell embedding provides a foundation model for cell biology. Nature 1\u20139 (2026). URL https:\/\/www.nature.com\/articles\/s41586-026-10689-z."},{"key":"2026091211100999000_2026.09.08.749477v1.61","doi-asserted-by":"crossref","unstructured":"Wang, C. et al. X-cell: Scaling causal perturbation prediction across diverse cellular contexts via diffusion language models. bioRxiv 2026.03.18.712807 (2026). URL https:\/\/www.biorxiv.org\/content\/10.64898\/2026.03.18.712807v1.","DOI":"10.64898\/2026.03.18.712807"},{"key":"2026091211100999000_2026.09.08.749477v1.62","unstructured":"Pao-Huang, P. , Qiu, X. & Ermon, S. Generative modeling with flux matching. arXiv (2026). URL https:\/\/arxiv.org\/pdf\/2605.07319."}],"container-title":[],"original-title":[],"link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.64898\/2026.09.08.749477","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,9,12]],"date-time":"2026-09-12T18:10:25Z","timestamp":1789236625000},"score":1,"resource":{"primary":{"URL":"http:\/\/biorxiv.org\/lookup\/doi\/10.64898\/2026.09.08.749477"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9,10]]},"references-count":62,"URL":"https:\/\/doi.org\/10.64898\/2026.09.08.749477","relation":{},"subject":[],"published":{"date-parts":[[2026,9,10]]},"subtype":"preprint"}}