Clinical

TrialMatch-NLP

TrialMatch-NLP is a SYNTHETIC/DEMO clinical model for patient-to-clinical-trial matching. Outputs are deterministic demo values — NOT for clinical use. Trained (synthetically) on de-identified multi-modal data.

Multi-modal EMBL-Synthetic License: Apache-2.0 v2.2.8 RTG 4.0 DL 207,743
SYNTHETIC DEMOAll outputs are deterministic demo data — not for clinical use, not real medical inference.

Metrics

MetricValueSplit
Recall@100.871Synthetic hold-out
Calibration (ECE)0.025Synthetic
InputsOutputs
Patient record + criteriaRanked eligible trials

Run a synthetic prediction

Returns a deterministic synthetic result — clearly labeled, not for clinical use.

API integration

Call this model from your pipeline (demo — no real key required):

import requests
r = requests.post("https://api.biofoundry.ai/v1/models/bf-043-trialmatch-nlp/predict",
    headers={"Authorization": "Bearer <API_KEY>"},
    json={"input": "<patient record + criteria>"})
print(r.json())  # -> synthetic demo prediction

See the API reference · JSON: /api/models/bf-043-trialmatch-nlp

Version history

VersionStatusNotes
v2.2.0ArchivedSynthetic release note 0
v2.2.1ArchivedSynthetic release note 1
v2.2.2ArchivedSynthetic release note 2
v2.2.3ArchivedSynthetic release note 3
v2.2.4ArchivedSynthetic release note 4
v2.2.5ArchivedSynthetic release note 5
v2.2.6ArchivedSynthetic release note 6
v2.2.7ArchivedSynthetic release note 7
v2.2.8CurrentSynthetic release note 8

Community contributions: 77 synthetic contributors have submitted evaluations and fine-tunes.