Tahoe-100M: From Open Preprint to Peer Review in Cell
The Tahoe-100M cancer-cell atlas became a benchmark AI dataset while still an unreviewed preprint; the peer-reviewed paper appeared in Cell roughly 19 months later.
In February 2025, the team now called Tahoe (previously Vevo Therapeutics) released the Tahoe-100M atlas: gene-activity profiles of 95.6 million cells from 50 cancer cell lines (laboratory cultures) treated with 379 drugs and drug candidates. That is 31 times more than the previous benchmark single-cell drug screen. The peer-reviewed paper appeared in Cell on September 17, 2026, roughly 19 months after the preprint.
The central challenge at this scale: the more experiments you run, the harder it becomes to separate a drug's real effect from random differences between batches (different days, reagents, technicians). Conventionally, each cell line is tested separately, and these technical differences accumulate. The authors pooled cells from all lines into a single vessel (a "cell village") and treated the mixture with one drug at a time. After RNA sequencing, each cell was assigned back to its line of origin using the cell's own genetic markers. Because all lines received identical treatment, the remaining differences reflect each line's genuine drug response. Re-running the same batch yielded 97 to 98 percent concordance. This collection method passed peer review.
Between the preprint and publication, the dataset was downloaded more than 600,000 times and became a benchmark for AI models that predict a cell line's drug response without running additional experiments.
The data come from laboratory cultures, collected over 24 hours of treatment. The study does not establish what these gene-activity changes mean for treating patients.
The Tahoe-100M publication timeline — open preprint February 2025, peer review September 2026, 600K+ downloads in between — provides a grounded, non-abstract illustration of how open data creates value before peer review. This directly anchors the Experiment angle: Experiment is a research crowdfunding platform where backers receive data, protocols, and papers, and the Open Access badge formalises that commitment. The news item gives readers a real example against which to measure any project's openness promise.