Nikhil Thaker: Excited to Share Our New Work on AI-Powered Oncology Clinical Trial Matching
Nikhil Thaker/LinkedIn

Nikhil Thaker: Excited to Share Our New Work on AI-Powered Oncology Clinical Trial Matching

Nikhil Thaker, Medical Director of Oncology Research at Capital Health (US), shared a post on LinkedIn:

“Excited to share our new work on AI-powered oncology clinical trial matching! Exceptional work led by the incredibly versatile Arturo LoAIza-Bonilla and amazing co-authors Connor Yost, Selin Kurnaz,  Ertugrul Tuysuz, Dersu Giritlioglu, Juan Pablo Noel Meza.

In this prospective evaluation of 3,804 patients, we studied a neuro-symbolic, multi-agent AI platform that combines LLM-based agents, an oncology-specific knowledge graph, real-time recommendation, and clinician oversight. The goal: make clinical trial matching more accurate, efficient, and scalable in real-world oncology practice.

A few key findings:

  • F1 score of 0.82, compared with 0.47 for GPT-4 zero-shot and 0.67 for GPT-4 chain-of-thought
  • Screening time reduced from about 120 minutes manually to roughly 30 minutes total with AI assistance
  • The system processed 157,367 clinical pages and generated 17,912 oncologist-confirmed matches
  • Performance gaps across demographic subgroups remained below the pre-specified 10-point fairness alert threshold

What makes this especially promising is the architecture: rather than relying on LLMs alone, the system uses a knowledge graph and deterministic eligibility logic to improve auditability, reduce hallucinations, and support safer clinical deployment with a human-in-the-loop.

This kind of hybrid, clinically grounded AI may represent an important path forward for responsible deployment in oncology, with groups like Massive Bio leading the way to improving trial access while preserving transparency, clinician oversight, and attention to equity.

Proud to be part of this work with a terrific team of collaborators! Thanks, Arturo LoAIza-Bonilla!”

Title: Transforming oncology clinical trial matching through neuro-symbolic, multi-agent AI and an oncology-specific knowledge graph: a prospective evaluation in 3804 patients

Authors: Arturo LoAIza-Bonilla, Connor Yost, Selin Kurnaz,  Ertugrul Tuysuz, Nikhil Thaker, Dersu Giritlioglu, Juan Pablo Noel Meza

Read the article.

Nikhil Thaker

Other articles featuring Nikhil Thaker on OncoDaily.