Abstract B016: Multi-Agent Framework for Deep Research in Cancer Immunogenomics via TCR Datasets and Scientific Literature Search
Bibliographic record
Abstract
Abstract Recent advances in cancer immunogenomics and the emergence of agentic artificial intelligence highlight the urgent need and opportunity for analytical frameworks capable of integrating large-scale immune repertoire datasets with scientific literature to accelerate discovery and hypothesis generation. We propose a novel, conceptual multi-agent framework designed to iteratively query and analyze publicly available T-cell receptor (TCR) databases—such as VDJdb, and TCRdb—alongside automated literature retrieval, enabling deeper exploration of immunological signatures associated with adverse events in cancer treatments. Our approach utilizes LangGraph to orchestrate specialized agents built with PydanticAI to ensure structured and robust interactions with diverse data sources. Within this system, an orchestrator agent manages user queries and delegates tasks to dedicated agents responsible for immune repertoire data extraction, scientific literature mining, and integrative analysis. We have implemented a version of the Data Agent, which queries immune repertoire datasets using natural language to generate analyses and visualizations. This initial agent can perform complex diversity analyses, identification of unique clonotypes, and statistical comparisons between patient groups experiencing different clinical outcomes. We are finalizing iterative refinement and integration with external scientific knowledge sources. In the proposed framework, iterative reasoning cycles will allow agents to dynamically refine queries and hypotheses based on intermediate findings, while observability tools will continuously monitor agent performance, interactions, and system health to ensure reliability and transparency. Illustrative use cases highlight the framework's potential applications, including: (1) detection of immune-related adverse events (irAEs) during immunotherapy by identifying characteristic TCR motifs, and (2) identification of biomarkers predictive of adverse events in combination therapies through comparative TCR analyses. Structured outputs from the system encompass detailed reports and visual summaries, clearly mapping the analytical journey from raw data queries through iterative reasoning to final hypotheses, explicitly highlighting the chain of thought, responsibilities, and contributions of each agent involved. Designed with flexibility and scalability in mind, our framework supports deployment locally as a Python application, within Docker environments, or exposed via an API, facilitating seamless integration into diverse research workflows. Furthermore, we intend to release the framework as an open-source project to encourage collaboration and community-driven advancements. Citation Format: Samuel Torres-Florez, Ramanandan Prabhakaran, Rajat Mohindra, Adrian B. Roth, Cameron MacPherson. Multi-Agent Framework for Deep Research in Cancer Immunogenomics via TCR Datasets and Scientific Literature Search [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Artificial Intelligence and Machine Learning; 2025 Jul 10-12; Montreal, QC, Canada. Philadelphia (PA): AACR; Clin Cancer Res 2025;31(13_Suppl):Abstract nr B016.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.005 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one teacher head, not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".