Clinical Trial Prediction Challenge

With a staggering 90% failure rate in oncology trials, the need for innovation has never been greater.
While AI is transforming many fields, its potential to radically improve clinical trial outcomes is still untapped.

As former participants in CASP (worldwide experiment for protein structure prediction), we know that without proper benchmarking, providing real-life actionable feedback—it is nearly impossible for innovators to navigate technology development.

That’s why we’re launching CAST—Critical ASsessment of techniques for Trial outcome prediction. Just as CASP paved the way for AlphaFold, we believe CAST will spark the breakthrough that will redefine how clinical trials are conducted.
Our bold goal? A 90% success rate in oncology clinical trials.

If you’re ready for the challenge, submit your prediction and join us in reshaping the future of clinical trials.

Based on clinical trial design, participants are to develop models that predict trial outcomes. This edition focuses on clinical efficacy: predict the percentage of patients meeting primary and secondary endpoints, along with a list of response and resistance biomarkers. Don’t worry—there’s no need to reveal your proprietary methods or "secret sauce."

How It Works

Richard Feynman once said, "You can't foresee outcomes unless you fully understand how the system works." CAST challenges companies, startups, academics, and innovators to develop technology that accurately predicts the efficacy in phase 2 oncology trials. Expected byproduct—a deeper understanding of how the system (cancer) works.

The Data Behind the Challenge

The CASP challenge centers on protein structure prediction, while the CAST competition focuses on forecasting clinical trial outcomes. Just as CASP has driven progress in computational structural biology, CAST aims to ignite the same level of innovation in improving clinical response rates and advancing the future of medicine.

Participants will assess real, ongoing phase 2 clinical trial data, including treatment specifics and drug targets. You'll analyze patient cohorts and predict which groups will respond to the treatment—and why.

There is no registration needed. You can upload your predictions by clicking the Submit Now! button and filling out a short questionnaire. Submissions should include predicted response rates, key biomarkers, and reasoning behind your predictions.

Data Access

You will have direct access to a list of ongoing phase 2 clinical trials. Each trial will have a summary, including treatment specifics and drug targets. Simply click on the trial to get more detailed information and begin analyzing.

Evaluation

Timeline

Submissions will be evaluated based on the accuracy of predictions compared to the actual published trial outcomes. How close were your predictions to the real results? Your model’s accuracy and insights into biomarkers will be key factors. No details about your approach are required. Only the top three performing teams will be named and invited to present their approach.

March 15th, 2025: Trials selected for the competition
June 30th, 2025: Predictions must be submitted
October 15th, 2025: Results will be revealed, and winners will be announced
December 15th, 2025: Celebration event in Boston.