Origin

Origin Bio is a computational oncology company that uses deep learning on real patient tumor tissue to predict which patients are likely to respond to a cancer drug. The stated aim is to identify responders before a trial is designed, which the company argues would raise success rates, especially in Phase 2. The approach starts with formalin fixed tumor blocks that arrive from hospital and biobank partners with anonymized clinical annotations such as tumor type, grade, stage and treatment history. Pathologist annotated regions are sectioned onto slides, then profiled with spatial transcriptomics at subcellular resolution alongside H&E imaging from the same slide. From each sample the company derives spatially resolved cell states, cell to cell interactions, pathway enrichment and biomarkers, and it trains models on the growing dataset to generalize across the variation between and within tumors. A web workspace lists runs on uploaded Xenium slides and shows results such as predicted cell counts, for example more than 57,000 predicted cells in a lung sample. The site is aimed at drug developers and oncology researchers, and access is requested through a Get Access button. Advisors named on the site include researchers from MIT and the Broad Institute, UCSF and the University of Illinois, and Y Combinator is listed among the investors. No pricing is published.