August 2026
OPTIONS-ACTIVATE launches
A randomized study testing whether proactive AI trial suggestions help more patients reach clinical trials.
MatchMiner is a suite of tools that help cancer centers with clinical trial enrollment. It has been built and used at Dana-Farber Cancer Institute since 2016.
MatchMiner Genomics
EGFR p.L858R Lung adenocarcinoma
Matches 24-117, arm A
MatchMiner-AI
Summary from notes Metastatic lung adenocarcinoma, EGFR L858R. Progressed on osimertinib. Good performance status, adequate organ function.
MatchMiner Dashboard
| Trial | Status | Open slots |
|---|---|---|
| 24-117 | Open | 3 |
| 25-032 | Open | 1 |
| 25-210 | Upcoming | 0 |
Clinical trials are how new cancer treatments reach patients. Yet fewer than 1 in 10 adults with cancer take part,1 and many trials miss their enrollment goals or never finish.2,3
Oncology trials are complex, with multiple arms and narrowly defined eligibility, often including specific treatment and biomarker requirements.4 Working through those criteria for every patient takes time that oncologists and trial teams rarely have. MatchMiner does that work at scale, checking patients against every open trial so oncologists and trial teams see the trials each patient may qualify for.
Genomic matching
Matching patients to the specific genomic requirements of precision medicine trials is a complex task.
Trial eligibility is translated into a structured format and entered into a matching engine that compares each trial to patients’ tumor sequencing results. It can find trials for a specific patient, or patients for a specific trial.
In clinical use at Dana-Farber since 2016.
AI matching
Details that decide eligibility, like prior treatment, are often found only in clinical notes.
Language models read a patient’s clinical notes and match them against all open trials, not only genomically driven ones. Oncologists review the suggested trials, each with a plain-language summary.
Launched for Dana-Farber providers in February 2026.
Trial operations
Trial teams track open trials, slots, and accrual across several systems and a lot of email.
One view for trial management: open slots, enrollment, and milestones. Details come from ClinicalTrials.gov and the center’s clinical trial management system (CTMS), and teams add eligibility notes.
In production at Dana-Farber across several trial teams.
The matching engine, the web application, and MatchMiner-AI’s model weights, inference code, and synthetic training data are all public. Any cancer center can run them inside its own network and see exactly how a match was made. The MatchMiner Dashboard code is due for release in winter 2026–27.
Other centers already build on it. Princess Margaret Cancer Centre in Toronto leads CTIMS, an open-source trial matching system that uses the MatchMiner matching engine. See who uses MatchMiner.
Models and synthetic data
Documentation
MatchMiner has run at Dana-Farber since 2016. Providers now open it from the patient’s record in Epic, Dana-Farber’s electronic health record.
Dana-Farber also runs randomized studies to learn how best to get matches in front of oncologists, and changes the design based on the results.
Randomized trial, 2023–2024
When AI detected that a patient’s cancer was progressing, their oncologist got an email about trial matches from MatchMiner Genomics. The AI reliably spotted patients likely to need a new treatment, but the emails alone didn’t raise enrollment.
Randomized trial, recruiting since 2026
OPTIONS-ACTIVATE sends oncologists a ranked list of trials from MatchMiner-AI when progression is detected, and measures the share of new treatments that are clinical trials.
August 2026
A randomized study testing whether proactive AI trial suggestions help more patients reach clinical trials.
June 2026
Recognition for matching patients to trials from the full medical record, with openly shared models.
February 2026
Providers can now look up AI-suggested clinical trials for their patients.
MatchMiner Genomics and MatchMiner-AI are free to use and run behind your institution’s firewall. Find out what each one needs in data, infrastructure, and people.