Keiji AI LogoKeiji AI

Research · Nature Communications 2024

TrialGPT

Research on patient-to-trial matching. TrialGPT matches patients to clinical trials in three steps: it retrieves candidate trials, predicts eligibility for each criterion with an explanation, and scores each trial.

Who created TrialGPT?

TrialGPT is a National Institutes of Health project, hosted by the National Library of Medicine's NCBI, with Zhiyong Lu as senior author. Its authors include Keiji AI co-founder Zifeng Wang and Jimeng Sun, Keiji AI's CEO and co-founder, who is listed at the University of Illinois Urbana-Champaign.

What the paper found

  • Evaluated on three cohorts of 183 synthetic patients, with over 75,000 trial annotations.
  • Retrieval recalled over 90% of relevant trials using less than 6% of the collection.
  • Criterion-level accuracy was 87.3% on 1,015 manually evaluated patient-criterion pairs, close to expert performance.
  • A user study found a 42.6% reduction in screening time.

How TrialGPT relates to TrialMind

It is research, not a TrialMind module or a separately sold product. Published methods like this one inform how TrialMind works; TrialMind is the product Keiji AI sells.

See how TrialMind handles this work: Clinical development and operations.

The paper

Qiao Jin, Zifeng Wang, Charalampos S. Floudas, Fangyuan Chen, Changlin Gong, Dara Bracken-Clarke, Elisabetta Xue, Yifan Yang, Jimeng Sun, Zhiyong Lu. Matching patients to clinical trials with large language models. Nature Communications, 2024.

Details checked against the published record on 2026-11-01.

Frequently asked questions

Who created TrialGPT?
TrialGPT is a National Institutes of Health project, hosted by the National Library of Medicine's NCBI, with Zhiyong Lu as senior author. Its authors include Keiji AI co-founder Zifeng Wang and Jimeng Sun, Keiji AI's CEO and co-founder, who is listed at the University of Illinois Urbana-Champaign.
Is TrialGPT part of TrialMind?
No. It is research, not a TrialMind module or a separately sold product. Published methods like this one inform how TrialMind works; TrialMind is the product Keiji AI sells.
What did the TrialGPT paper find?
Evaluated on three cohorts of 183 synthetic patients, with over 75,000 trial annotations. Retrieval recalled over 90% of relevant trials using less than 6% of the collection. Criterion-level accuracy was 87.3% on 1,015 manually evaluated patient-criterion pairs, close to expert performance. A user study found a 42.6% reduction in screening time.

All publications

Every paper sits under its primary research area.