Orakl Raises $34M, Touting up to 90% Clinical Trial Prediction Accuracy. Can It Hold?

Paris techbio Orakl Oncology has raised a $34 million Series A co-led by Norrsken VC and Singular. InfraVia, Serena, Bpifrance’s Patient Autonome Fund, teampact.ventures and Plug and Play also joined the oversubscribed round, bringing total funding to $50M.

Orakl builds “virtual patients”: real patients’ tumours paired with their medical history and deep biological data, so treatments can be tested on them at scale. On its website, the company says its model “prospectively predicts individual patient response to new drug candidates with 90% accuracy across 30+ clinical trials, outperforming every competing model.”

Its published evidence is so far rather thin. An April 2026 preprint, not yet peer-reviewed, describes Orakl’s SCOPE model. It combines organoid drug screening with clinical prognostic modelling to predict median progression-free survival and response rates for each trial arm. Validated against 32 arms from 23 published colorectal and pancreatic cancer trials, predicted median PFS closely matched reported results (R² = 0.85, mean error 0.82 months), and response-rate predictions reached R² = 0.71. Combining organoid and clinical data performed significantly better than either alone. The model identified the better arm in 7 of 8 head-to-head comparisons (88%).

That is a serious attempt to forecast trial results, but it is not the claim on the website. The preprint predicts outcomes for trial arms, not individual patients. Its validation used trials whose results were already published, not prospective ones, and it reports no 90% accuracy figure or comparison against competing models. The evidence base is also small. The model was trained on 54 treatment lines from 52 patients and 9 compounds, all in gastrointestinal cancers, and the ranking result rests on eight comparisons. Its one forward-looking test, on the pan-RAS inhibitor daraxonrasib in pancreatic cancer, predicted superiority over chemotherapy, which the authors describe as consistent with emerging clinical data. The authors include Orakl’s three co-founders.

Orakl is not alone in this gap. Our 2026 watchlist of companies applying AI to de-risk clinical trials maps the field across trial design, control evidence, and digital twins, and operational execution. The recurring problem is the same: most headline numbers come from the companies themselves, and independent validation remains the open question.

For Orakl, that test is already under way. In July, it took exclusive rights to four undisclosed Sanofi oncology assets -- it would be interesting to watch the outcome of the program.