MatchMiner-AI

Clinical AI-Powered

MatchMiner-AI is a next-generation clinical trial matching platform that uses large language models (LLMs) to match patients to all clinical trials — not just genomically driven ones — based on their full medical record. The platform processes unstructured clinical notes, extracts key patient attributes, and generates plain-language patient summaries to identify the best-fit trials. MatchMiner-AI launched at Dana-Farber Cancer Institute in February 2026.

Key Capabilities

LLM-Powered Matching

Uses large language models to interpret complex trial eligibility criteria and match them against a patient's full clinical picture — diagnosis, history, labs, treatments, and more.

Unstructured Note Processing

Extracts structured clinical attributes from free-text physician notes, pathology reports, and other unstructured data sources that traditional rule-based systems cannot parse.

Patient Summaries

Automatically generates concise, plain-language summaries of a patient's clinical profile, helping oncologists quickly assess trial fit without manually reviewing the full record.

Broad Trial Coverage

Matches patients across all clinical trials — including those based on histology, prior treatment, performance status, and comorbidities — going far beyond genomic eligibility alone.

Clinical Integration

Integrated into the clinical workflow at DFCI, ingesting patient data from the electronic medical record and surfacing trial matches at the point of care.

Clinician Review

Each match surfaces a patient summary and a trial summary side by side, giving clinicians the context they need to evaluate fit and make informed decisions.

Beyond Genomic Matching

Traditional trial matching platforms — including MatchMiner Genomics — focus on genomic eligibility: matching patients based on specific mutations, fusions, or copy number alterations. This works well for genomically driven trials, but most clinical trials have eligibility criteria that go far beyond genomics.

MatchMiner-AI addresses this gap by reading the patient’s full medical record. It can interpret criteria like prior lines of therapy, specific diagnoses, organ function, performance status, and treatment history — criteria that are often buried in unstructured clinical notes and impossible to match with rule-based logic alone.

The result is broader, more complete trial coverage for every patient seen at DFCI.

Publication

Altreuter J, Trukhanov P, Paul MA, Hassett MJ, Riaz IB, Mallaber E, Klein HR, Gungor G, D'Eletto M, Van Nostrand SC, Provencher J, Mazor T, Cerami E, Kehl KL, et al. · arXiv preprint, 2024

GitHub →