Orakl Oncology raises $34 million to scale AI cancer predictions

Norrsken VC announced an oversubscribed $34 million Series A for Orakl Oncology on November 12, 2026, bringing the AI drug-development company’s total funding to $50 million. The financing puts a patient-derived laboratory model under a broader commercial test: whether its predictions can support more drug programs, more cancer types and longer pharmaceutical partnerships.

What does the funding buy?

Norrsken co-led the round with Singular. New investors included InfraVia, Serena, the Patient Autonome Fund managed by Bpifrance, teampact.ventures and Plug and Play. Existing backers Verve Ventures, Speedinvest and HCVC also participated.

Orakl says the proceeds will support several expansions:

• Expand its dataset and AI research team.
• Increase data generation fivefold.
• Establish teams in the US and UK.
• Deploy predictions across more than 10 drug programs.
• Move beyond colorectal and pancreatic cancers into lung and gastric cancers.
• Discover and validate drug co-targets, or targets considered for combination therapies, from patient biology and experimental screens.

The announcement gives no delivery schedule, hiring target or cash allocation. Fivefold data growth, new country teams and a wider disease footprint therefore remain plans rather than completed capabilities.

Takeaway: The money funds a wider reach, not a finished expansion.

How does Orakl build its prediction data?

Founded in 2023 as a spin-out of Gustave Roussy, Orakl grows patient-derived tumor organoids. These are laboratory models of individual patients’ cancers, and the company screens them against drugs.

Orakl’s platform overview describes results enriched with genomic, transcriptomic and phenotypic data and anchored to longitudinal clinical histories. Genomic data describes DNA, transcriptomic data tracks RNA activity, and phenotypic data captures observable traits. Linking those records across time creates a multimodal dataset, meaning the platform combines several biological and clinical data types.

Orakl divides its biopharma products between two systems:

• O-Predict is designed to anticipate trial endpoints, including tumor response, progression-free survival and overall survival. It also identifies biomarkers, or measurable biological signs, that distinguish likely responders, assesses drug combinations and compares benefit-risk profiles.
• O-Target focuses on therapeutic-target discovery.

The systems address different development questions. O-Predict is designed to estimate likely patient and trial outcomes, while O-Target searches for new therapeutic targets. The announcement gives no prospective accuracy result for O-Target.

Norrsken calls the resulting patient cohorts “virtual cancer patients.” That label does not make the underlying measurements synthetic: Norrsken said they must come from patient samples, laboratory infrastructure and hospital partnerships rather than existing databases. Each experiment adds proprietary data connecting patient biology, clinical information and observed drug responses.

The model begins with a clinical sample rather than a row in an existing database. Orakl screens the resulting organoid, joins the response to genomic, transcriptomic, phenotypic and clinical records, and adds the result to its proprietary dataset. The source does not specify how many samples, screens or compounds the expanded dataset will contain.

Takeaway: Orakl’s advantage depends on proprietary patient data, while its predictive performance still needs independent testing.

What does the SCOPE preprint show?

Orakl reports 90% accuracy for prospective predictions of individual patient responses across more than 30 clinical trials. The announcement does not define the accuracy calculation or disclose trial-by-trial denominators. That leaves the 90% headline without enough context for an independent comparison.

The 90% claim and the preprint’s R² scores use different summaries, and the announcement does not connect them through a shared denominator or protocol. A separate April 2026 preprint co-authored by Orakl staff provides arm-level results for SCOPE, the Screening-to-Clinical Outcome Prediction Engine.

The preprint is marked as early-stage research and may not yet have been peer reviewed. The researchers trained SCOPE on 54 treatment lines, meaning courses of therapy, from 52 patients with metastatic colorectal or pancreatic cancer. They used matched clinical data and organoid screens covering nine compounds.

For predictions, the system generated synthetic patient profiles from published eligibility criteria. It matched those profiles to a biobank, or stored collection, of 81 patient-derived organoid lines. External validation covered 32 arms, or patient groups, from 23 published trials.

For predicted median progression-free survival, the time patients live without disease progression, the authors reported an R² of 0.85 and a mean absolute error of 0.82 months. R², the coefficient of determination, measures fit and reaches 1 for a perfect match; mean absolute error gives the average size of a prediction miss.

The authors also report that combining organoid and clinical data significantly outperformed either data type alone. The source summary provides no effect size or confidence interval for that improvement. For objective response rate, the share of patients with a measured tumor response, R² was 0.71. Mean absolute error was 7.3 percentage points.

SCOPE selected the superior treatment arm in seven of eight head-to-head comparisons, according to the preprint. The document does not specify the selection margin or uncertainty around those comparisons.

The preprint also describes a prospective application to daraxonrasib before Phase III, the late-stage clinical testing phase, data became available. SCOPE predicted superiority over standard chemotherapy in KRAS-mutant metastatic pancreatic cancer, meaning pancreatic cancer that has spread and carries an altered KRAS gene. The source gives no patient count, prediction date or supporting outcome data for that application.

The detailed validation still concerns arm-level outputs from published trials, not outcomes assigned prospectively by SCOPE. The study therefore tests predictions against historical trial results; the daraxonrasib example describes a prospective use without reporting a prospective trial result.

Takeaway: SCOPE reports strong aggregate results, but its clinical claims remain preliminary.

What does commercial traction actually prove?

Just over a year after commercial launch, Norrsken reports that 18 of the world’s 20 largest pharmaceutical companies sit in Orakl’s commercial pipeline. Norrsken said the figure reflects “pipeline participation,” not signed customers, although the first pharmaceutical engagements have become multi-year collaborations.

More than 15 additional cancer centers across Europe, North America and Asia are in contract negotiations. Those relationships indicate commercial interest, but they do not establish a recurring revenue base. Pipeline participation can precede a contract, and negotiations can end without an agreement. The announcement provides no revenue, contract value or deployment count.

Norrsken reports that Orakl has more than 40 employees, and 30% hold PhDs. Co-founders CEO Fanny Jaulin, CTO Gustave Ronteix and COO Diane-Laure Pagès lead the company. Norrsken says their backgrounds span cancer biology, AI, data science, oncology and hospital partnerships.

The source does not report how many active programs carry live predictions. Nor does it provide renewal or expansion figures for the first multi-year collaborations. The commercial metric therefore shows breadth, not the scale of successful deployments.

The next phase will test whether Orakl can extend its patient-derived data model into lung and gastric cancers while converting its broad pharmaceutical pipeline into more multi-year drug-development collaborations. Takeaway: The $34 million buys scale; peer review, prospective results and signed contracts will determine whether the company can convert pipeline breadth into repeatable, multi-year development programs.