The tools
What’s the difference between MatchMiner Genomics and MatchMiner-AI?
MatchMiner Genomics matches patients to precision medicine trials using structured sequencing results and trial criteria that curators write in CTML. MatchMiner-AI uses language models to read clinical notes and match patients against all open trials, including ones with no genomic criteria. Dana-Farber runs both.
Does MatchMiner-AI send patient data outside my institution?
Not by default. MatchMiner-AI can run its models on your own hardware. It can also connect to a model server you choose, in your network or outside it. Clinical data should only be sent to an endpoint approved for patient data at your institution.
What data was MatchMiner-AI trained on?
Synthetic medical records only. That’s why the model weights and training data can be shared. The datasets are on Hugging Face.
Can I use MatchMiner Dashboard?
Not yet. MatchMiner Dashboard is in production at Dana-Farber, and its code is due for release as open source in winter 2026–27.
Setting up MatchMiner Genomics
Can I set up MatchMiner Genomics at my institution?
Yes. MatchMiner Genomics is open source and runs entirely inside your institution’s network, so patient data stays under your own security and HIPAA controls. The documentation covers deployment and a demo setup. See Get started for an overview.
What data do I need?
The MatchEngine and web interface need structured patient genomic and clinical data, and trial eligibility criteria written in CTML.
What hardware and software do I need?
A small instance runs on a laptop with 4 CPUs and 4 GB of memory. For comparison, Dana-Farber’s deployment (as of 2026, about 80,000 clinical samples, 700 trials, 600 saved filters, and 500 users) uses 45 CPU cores and 128 GB of memory.
What staff do I need?
We suggest at least two scientists to curate and review CTML documents, and at least one software engineer to set up MatchMiner. The engineer builds the data ingestion pipelines and integrates MatchMiner with institutional systems such as authentication.
How do I write trial eligibility in CTML?
You can write CTML by hand in a text editor, or use the CTML editor built by the CTIMS project at Princess Margaret Cancer Centre.
Where are the documentation and source code?
Documentation, including example CTML and example patient data, is on GitBook. Source code for the MatchEngine, API, and web interface is on GitHub.
Setting up MatchMiner-AI
How do I set up MatchMiner-AI?
Get started covers what it needs and how to install it.
Using MatchMiner
Who uses MatchMiner?
Clinicians, clinical research staff, and trial investigators use MatchMiner to find trial matches. Trial teams use MatchMiner Dashboard to track their trial portfolios.
What has MatchMiner Genomics’ impact been at Dana-Farber?
In January 2022, Dana-Farber had about 170 open precision medicine trials in MatchMiner. Patients matched to 6 trials on average, and 84% of patients had at least one match. By then, more than 220 patients had enrolled on a trial as a result of a MatchMiner match.
Getting help
I have a question that isn’t answered here.
I found a bug.
For bugs in the open-source code, open an issue in the relevant GitHub repository. For anything else, contact the team.