trialgpt-matching

name: trialgpt-matching description: Trial shortlist keywords: - retrieval - ranking - ClinicalTrials - patient-profile measurable_outcome: Produce ≥5 ranked trials (when available) with rationale + missing-data notes within 3 minutes of receiving a patient query. license: MIT metadata: version: "1.0.0" compatibility: - system: Python 3.9+ allowed-tools: - run_shell_command - read_file

TrialGPT Matching

Run the locally checked-out TrialGPT pipeline to retrieve, rank, and explain candidate trials for a patient before deeper eligibility review.

Core Capabilities

• Normalize oncology patient facts and trial criteria through an oncology-specific knowledge graph, parse eligibility text into criterion objects and symbolic eligibility constraints, assign specialized agents for evidence extraction, criteria harmonization, rule execution, discrepancy adjudication, and clinician review, record prospective patient-level evaluation metrics, and keep clinician override review in the final decision loop.
• Add neuro-symbolic multi-agent oncology trial matching patterns that use an oncology-specific knowledge graph for criterion-level reasoning, capture prospective evaluation metrics, and preserve patient-scale auditability for large clinical cohorts such as the reported 3804-patient evaluation.
• Use a neuro-symbolic multi-agent oncology trial-matching pattern grounded in an oncology-specific knowledge graph: reason at the eligibility-criterion level, carry patient cohort-scale prospective evaluation context from the 3804-patient study, resolve conflicting agent judgments, and preserve audit trails for human trial-navigation review.
• For prospective oncology trial matching, coordinate neuro-symbolic, multi-agent review over an oncology-specific knowledge graph; attach criterion-level evidence to each inclusion/exclusion judgment, document patient-scale validation context from the 3804-patient evaluation, reconcile agent conflicts before final recommendations, and require human review checkpoints before matches are acted on.
• Use neuro-symbolic multi-agent oncology trial matching patterns that ground patient facts and trial criteria in an oncology-specific knowledge graph, run criterion-level symbolic reasoning, preserve prospective patient-scale evaluation context, resolve conflicts among agents, and retain audit trails for each match recommendation.
• Incorporate neuro-symbolic, multi-agent oncology trial matching grounded in an oncology-specific knowledge graph, with criterion-level eligibility reasoning, prospective cohort evaluation context, audit trails, and human review for ambiguous inclusion or exclusion criteria.
• Coordinate multi-agent criteria extraction for oncology patient-trial recommendations by grounding extracted patient facts and trial criteria in an oncology-specific knowledge graph, producing criterion-level explanations, capturing prospective evaluation metrics, and requiring human review before recommendations are acted on.
• Ground neuro-symbolic multi-agent oncology trial matching in an oncology-specific knowledge graph for large patient cohorts, with criterion-level eligibility reasoning, prospective evaluation metric capture, confidence scoring, and audit trails for each patient-trial decision.
• Decompose oncology trial eligibility criteria across neuro-symbolic agents grounded in an oncology-specific knowledge graph, preserve prospective cohort evaluation context from the 3804-patient study, score match confidence, and route eligibility conflicts or low-confidence decisions to human review.
• Apply neuro-symbolic, multi-agent oncology trial matching for large patient cohorts by grounding criteria in oncology-specific knowledge graphs, reasoning at the eligibility-criterion level, preserving prospective evaluation context, assigning confidence scores, and routing results through human review.
• For oncology trial matching informed by the 2026 prospective evaluation in 3804 patients, coordinate neuro-symbolic agents over an oncology-specific knowledge graph to reason at the eligibility-criterion level, preserve provenance for every match decision, track prospective evaluation metrics, and escalate uncertain criteria to human reviewers.
• Implement neuro-symbolic, multi-agent oncology trial matching grounded in an oncology-specific knowledge graph, with criterion-level evidence packets, prospective evaluation metrics, and human-in-the-loop adjudication informed by the 2026 prospective evaluation in 3804 patients.
• Separate eligibility parsing, patient evidence extraction, symbolic rule execution, discrepancy review, and site-level audit metrics for prospective oncology trial-matching workflows that use neuro-symbolic multi-agent screening with an oncology-specific knowledge graph.
• Apply a neuro-symbolic multi-agent trial-matching pattern for high-volume oncology workflows: ground retrieval and ranking in an oncology-specific knowledge graph, reason at the eligibility-criterion level, arbitrate patient-trial eligibility decisions, attach confidence scores, preserve prospective validation context, and route matches for clinician review.
• Incorporate prospective oncology trial matching patterns from neuro-symbolic multi-agent systems and oncology knowledge graphs: criterion-level reasoning, KG-backed eligibility checks, patient-scale evaluation context, confidence scoring, and oncologist review.
• Support prospective oncology trial matching for large real-world cohorts with neuro-symbolic, multi-agent workflows grounded in an oncology-specific knowledge graph, including criterion-level matching, evidence provenance, unresolved eligibility gaps, and human review flags.
• Apply prospective oncology trial matching patterns from large patient cohorts: neuro-symbolic multi-agent review, oncology knowledge graph grounding, criterion-level evidence extraction, unresolved-data notes, and clinician validation for each screening run.
• Use neuro-symbolic multi-agent trial matching patterns with oncology knowledge graph grounding, criterion decomposition, patient-trial matching across large prospective cohorts, contradiction handling across patient facts and eligibility criteria, and human-reviewable rationales linking evidence to eligibility decisions.
• Combine an oncology-specific knowledge graph with criterion-level eligibility reasoning and neuro-symbolic multi-agent review for high-volume prospective screening, routing patient-trial matches and unresolved eligibility gaps to human reviewers as in the 2026 prospective oncology evaluation in 3804 patients.
• Run criterion-level symbolic checks grounded in oncology knowledge graph context, score match confidence, and emit clinician-facing explanations and audit outputs for prospective cohort oncology trial screening.
• For prospective oncology trial matching, decompose eligibility into criterion-level checks, normalize biomarkers and regimens through the oncology knowledge graph, compare multi-agent judgments for disagreement review, assign confidence scores, retain audit trails, and define prospective validation endpoints before clinician sign-off.
• For large prospective oncology cohorts, coordinate neuro-symbolic agents over an oncology-specific knowledge graph to produce criterion-level eligibility reasoning, patient-level audit trails, confidence scoring, and human-review queues for each candidate match.
• Use a neuro-symbolic multi-agent oncology matching pattern that combines oncology knowledge graph grounding, criterion-level evidence extraction, eligibility conflict resolution, confidence scoring, and prospective audit metrics from patient-scale evaluations.
• Parse eligibility at the criterion level, ground oncology concepts in a knowledge graph, track prospective evaluation metrics, resolve conflicts between matching agents, and retain audit trails for patient-trial recommendations.
• Add clinician review gates to prospective oncology trial-matching runs: coordinate neuro-symbolic multi-agent checks over an oncology-specific knowledge graph, preserve criterion-level audit trails, and report cohort-scale evaluation context before recommendations are acted on.
• Ground neuro-symbolic multi-agent oncology trial matching in ontology and knowledge-graph context, reason over each eligibility criterion, resolve eligibility conflicts across patient evidence and trial rules, document prospective patient-scale evaluation context, and route uncertain matches through human review workflows.
• For high-volume oncology patient screening, design prospective evaluations around neuro-symbolic multi-agent outputs grounded in an oncology-specific knowledge graph: capture criterion-level eligibility reasoning, confidence scores, unresolved criteria, and human-in-the-loop review status before any match is treated as actionable.
• For large patient cohorts, combine neuro-symbolic agents with an oncology-specific knowledge graph to evaluate each inclusion or exclusion criterion, record and resolve agent disagreements, preserve prospective evaluation context, and route unresolved patient-trial decisions to human review.
• Use a neuro-symbolic multi-agent oncology review pattern that grounds patient facts and trial criteria in an oncology-specific knowledge graph, assigns review roles for evidence extraction, criteria harmonization, symbolic criterion checks, discrepancy adjudication, and clinician review, tracks prospective evaluation metrics, and escalates ambiguous, missing, or conflicting eligibility evidence for human adjudication.
• For high-volume oncology cohorts such as the 3804-patient prospective evaluation, ground patient facts and eligibility criteria in an oncology-specific knowledge graph, have neuro-symbolic agents produce criterion-level eligibility judgments, attach confidence scores and prospective evaluation metric fields without inventing benchmark values, and require human review for low-confidence, conflicting, or actionable matches.
• For prospective oncology trial matching, configure neuro-symbolic multi-agent workflows around an oncology-specific knowledge graph, emit criterion-level explanations with confidence scoring, preserve audit trails for each patient-trial decision, and report prospective validation metrics without inventing benchmark values.

Inputs

• Patient summary (structured JSON or free text) with condition keywords.
• Optional filters: geography, phase, intervention, biomarker.
• Up-to-date ClinicalTrials.gov dump or API access.

Outputs

• Ranked trial table with NCT ID, title, score, and short justification.
• Parsed inclusion/exclusion text ready for downstream eligibility agents.
• Missing data checklist (e.g., "ECOG not provided").
• Prospective oncology screening packet using neuro-symbolic criteria parsing, oncology-specific knowledge graph context, multi-agent eligibility review, criterion-level evidence, confidence scoring, and human review flags for high-volume patient matching.

Workflow

1. Setup: cd repo && pip install -r requirements.txt (or reuse env).
2. Trial retrieval: Run TrialGPT retriever to pull candidate trials for the indication.
3. Criteria parsing: Convert eligibility blocks to structured criteria JSON.
4. Patient profiling: Summarize patient facts (labs, prior therapies, biomarkers).
5. Ranking: Execute TrialGPT ranking script to score each trial and emit explanations.
6. Handoff: Export ranked list + structured criteria for trial-eligibility-agent.

Guardrails

• Refresh ClinicalTrials.gov metadata regularly to avoid stale trials.
• Label scores as AI-generated suggestions pending clinician validation.
• Retain prompt/config metadata for audit trails.

References

• Detailed usage instructions and repo layout live in README.md.
• Coordinate with Skills/Clinical/Trial_Eligibility_Agent for criterion-level review.
• https://pubmed.ncbi.nlm.nih.gov/42004487/