Owkin

Owkin uses artificial intelligence to find the right treatment for every patient.

[Owkin @OwkinScience]

It's why guarding against overfitting sits at the core of how we build at Owkin: training on data from many hospitals, scanners and patient populations, and stress-testing on cohorts the models haven't seen before.

Read our 5-min guide to Overfitting: https://t.co/J0yruSOY1q

[Owkin @OwkinScience]

You spend years and millions training a radiology AI. It's brilliant in your hospital, on your scanners, with your patients. But then you roll it out anywhere else… and it doesn’t work. (1/5)

[Owkin @OwkinScience]

In healthcare, this looks like:
🔹 A cancer-detection model that "learns" a scanner's colour profile instead of the disease
🔹 A model trained at one hospital that stumbles elsewhere
🔹 Strong lab results that quietly collapse in deployment
(4/5)

[Owkin @OwkinScience]

Gene signature and pathway enrichment is now in K Pro.ssGSEA across 320 MOSAIC NSCLC patients, adenocarcinoma vs squamous. MYC targets up in the squamous tumors, consistent with 3q26.Scored, tested, plotted in three minutes.

👉 Watch it run

[Owkin @OwkinScience]

@JonLevyTLB talks to @TClozel on the Owkin x @Super_Human_Net podcast about social integration and longevity, and what it takes to build real community.

Watch the podcast → https://t.co/Jy3pHNRIZr

[Owkin @OwkinScience]

Learn more about Owkin’s vision here: https://t.co/Q5cfLtSB1w

[Owkin @OwkinScience]

Through the combined power of our MOSAIC dataset, AI Scientist K Pro, and wet lab we shaped OKN3201, our bispecific ADC for solid tumors.

[Owkin @OwkinScience]

This is how Owkin is pioneering biological artificial superintelligence:
- Patient data, built with leading hospitals
- K Pro in the hands of our team and top pharma R&D
- Our own wet labs to test and validate hypotheses
- Drug dev. programs from target ID to clinical trials

[Owkin @OwkinScience]

If you're working on drug R&D and want to see what this looks like in practice, get in touch with us through the link below. → https://t.co/V0YVzXOpzU

[Owkin @OwkinScience]

Novelty can feel risky when no one has measured it before. That is probably one of the reasons why drug R&D keeps crowding around a relatively small number of usual targets.

[Owkin @OwkinScience]

Breaking it takes data from outside that loop, new patient data that doesn’t come from behind a single pharma firewall, with enough depth and breadth to support new training. That's what we bring through our AI Scientist, K Pro.

[Owkin @OwkinScience]

We’re proud to announce K Pro has been nominated for the 2026 Prix Galien USA Awards under Best Digital Health Solution.

Thank you to the Prix Galien committee for recognizing K Pro's role in advancing biomedical research.

Learn more about K Pro here → https://t.co/XV8RMXap30

[Owkin @OwkinScience]

Standard stats (DESeq2), with the query and every parameter visible and editable — not a black box.

👉 Watch it run

[Owkin @OwkinScience]

Pseudobulk DEA shipped in K Pro, our AI Scientist. We pointed it at the public CELLxGENE breast cancer atlas with one ask: find cell-surface markers specific to malignant cells. 🧵👇

[Owkin @OwkinScience]

Then we changed the question with one line: "restrict to TNBC patients and raise the logFC cutoff to 2."

It re-ran the analysis on the new cohort and threshold. No new code.