Agentic Discovery Transfer Enables Cross-Cancer Discovery at Scale
Abstract
Cancer research generates tens of thousands of studies each year, yet whether a published finding holds in another cancer type or patient population is rarely tested, because doing so is labor-intensive and difficult to scale. We define this task as discovery transfer and present SMART, a multi-agent large language model framework for discovery transfer in single-cell RNA-sequencing (scRNA-seq) studies. Given a source manuscript and a target scRNA-seq dataset, SMART extracts major clinical and mechanistic findings, generates and executes analyses for the testable findings, and reports which findings transfer and suggests follow-up. Applied to 121 studies across six cancer types, SMART identified findings that transfer across cancers, including a lung cancer prognostic signature that we confirmed to be associated with survival in liver cancer, pancreatic cancer and pleural mesothelioma. SMART also generated new mechanistic hypotheses, demonstrating how agentic discovery transfer can extend published findings beyond their original setting.
Related
- Cross-modal transfer enables zero-shot prediction of drug responses across cell types
- Hierarchical Temporal Transformer for Cancer Grade Prediction and Cross-Cancer Transfer Learning from Pathology Reports
- Hierarchical temporal transformer for cancer grade prediction and cross cancer transfer learning from pathology reports
- Agentic AI enables cell segmentation-free cell-type inference and cell-type-guided cell segmentation in spatial transcriptomics
- APOD: reasoning-guided agentic population ordinary differential equation discovery for pharmacological digital twins
BibTeX
@misc{cheng2026agentic,
title = {{Agentic Discovery Transfer Enables Cross-Cancer Discovery at Scale}},
author = {Lingxin Cheng and Jack Shireman and Akash GOGATE and Maggie Stangis and Yujia Cai and Yixuan Xie and Parth Khatri and Mahua Dey and Christina Kendziorski},
howpublished = {bioRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.28.762506},
url = {https://www.biorxiv.org/content/10.64898/2026.10.28.762506v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Agentic Discovery Transfer Enables Cross-Cancer Discovery at Scale
AU - Lingxin Cheng
AU - Jack Shireman
AU - Akash GOGATE
AU - Maggie Stangis
AU - Yujia Cai
AU - Yixuan Xie
AU - Parth Khatri
AU - Mahua Dey
AU - Christina Kendziorski
PY - 2026
DA - 2026/10/28
PB - bioRxiv
DO - 10.64898/2026.10.28.762506
UR - https://www.biorxiv.org/content/10.64898/2026.10.28.762506v1
AB - Cancer research generates tens of thousands of studies each year, yet whether a published finding holds in another cancer type or patient population is rarely tested, because doing so is labor-intensive and difficult to scale. We define this task as discovery transfer and present SMART, a multi-agent large language model framework for discovery transfer in single-cell RNA-sequencing (scRNA-seq) studies. Given a source manuscript and a target scRNA-seq dataset, SMART extracts major clinical and mechanistic findings, generates and executes analyses for the testable findings, and reports which findings transfer and suggests follow-up. Applied to 121 studies across six cancer types, SMART identified findings that transfer across cancers, including a lung cancer prognostic signature that we confirmed to be associated with survival in liver cancer, pancreatic cancer and pleural mesothelioma. SMART also generated new mechanistic hypotheses, demonstrating how agentic discovery transfer can extend published findings beyond their original setting.
SN - 2692-8205
ER -