Large language models enable tumor-type classification and localization of cancers of unknown primary from genomic data

Sep 23, 2026

Uploaded by Mary Man

Liu et al. introduce OncoChat, a large language model that classifies 69 tumor types by integrating diverse genomic alterations to support accurate tumor-type localization. Built on genomic data from 158,836 targeted-panel sequenced tumors, OncoChat achieves micro-averaged PRAUC of 0.810, accuracy of 0.774, and F1 score of 0.756, exceeding baseline methods. For cancers of unknown primary (CUP), it correctly identifies 22 of 26 confirmed cases, and in larger CUP datasets predicted tumor types correlate with survival outcomes and coherent mutation profiles.

LanguageEnglish

CategoryHealthcare

FormatPDF

Pages20

Reading Time50 min

AccessFree to download

Highlights

In brief

Summary

Introduction

Article Large language models enable tumor-type classification and localization of cancers of unknown primary from genomic data Graphical abstract Highlights • OncoChat, a large language model, classifies tumor types from genomic alterations Authors Jilei Liu, Meng Yang, Yajing Bi, ..., Hongru Shen, Kexin Chen, Xiangchun Li Correspondence [chenkexin@tmu.edu.cn](chenkexin@tmu.edu.cn) (K.C.), [lixiangchun@tmu.edu.cn](lixiangchun@tmu.edu.cn) (X.L.) In brief Liu et al. develop OncoChat, a large language model that accurately classifies multiple cancer types, including cancers of unknown primary (CUP), from genomic data. OncoChat outperforms existing models, and its predictions for CUP cases correlate with clinical outcomes, offering a promising new tool for personalized cancer therapy. • OncoChat classifies known and unknown primary cancers with superior accuracy • Structural variant integration greatly improves tumor classification accuracy • OncoChat predictions are prognostic for survival in cancers of unknown primary Liu et al., 2025, Cell Reports Medicine 6, 102332 September 16, 2025 © 2025 The Authors. Published by Elsevier Inc. [https://doi.org/10.1016/j.xcrm.2025.102332](https://doi.org/10.1016/j.xcrm.2025.102332) ll ll OPEN ACCESS Article Large language models enable tumor-type classification and localization of cancers of unknown primary from genomic data Jilei Liu,1,3 Meng Yang,1,3 Yajing Bi,1,3 Junqing Zhang,1,3 Yichen Yang,1 Yang Li,1 Hongru Shen,1 Kexin Chen,2,* and Xiangchun Li1,4,* 1Key Laboratory of Cancer Prevention and Therapy, Tianjin Cancer Institute, Tianjin’s Clinical Research Center for Cancer, Tianjin Key Laboratory of Digestive Cancer, State Key Laboratory of Druggability Evaluation and Systematic Translational Medicine, National Clinical Research Center for Cancer, Tianjin Medical University Cancer Institute and Hospital, Tianjin Medical University, Tianjin 300060, China 2Department of Epidemiology and Biostatistics, Key Laboratory of Prevention and Control of Major Diseases in the Population Ministry of Education, Key Laboratory of Molecular Cancer Epidemiology of Tianjin, Tianjin’s Clinical Research Center for Cancer, National Clinical Research Center for Cancer, Key Laboratory of Cancer Prevention and Therapy, Tianjin Medical University Cancer Institute and Hospital, Tianjin Medical University, Tianjin 300060, China 3These authors contributed equally 4Lead contact *Correspondence: [chenkexin@tmu.edu.cn](chenkexin@tmu.edu.cn) (K.C.), [lixiangchun@tmu.edu.cn](lixiangchun@tmu.edu.cn) (X. L.) [https://doi.org/10.1016/j.xcrm.2025.102332](https://doi.org/10.1016/j.xcrm.2025.102332) SUMMARY Tumor-type classification is critical for effective cancer treatment, yet current methods based on genomic alterations lack flexibility and have limited performance. Here, we introduce OncoChat, an artificial intelligence (AI) model designed to classify 69 tumor types by integrating diverse genomic alterations. Developed on genomic data from 158,836 tumors sequenced with targeted cancer gene panels, OncoChat demonstrates superior performance, achieving a micro-averaged precision-recall area under the curve (PRAUC) of 0.810 (95% confidence interval [CI], 0.803–0.816), accuracy of 0.774, and an F1 score of 0.756, outperforming baseline methods. In a cancer of unknown primary (CUP) dataset of 26 cases whose types were subsequently confirmed, OncoChat correctly identified 22 cases. In two larger CUP datasets (n = 719 and 158), tumor types predicted by OncoChat were associated with survival outcomes and mutation profiles consistent with those of known tumor types. OncoChat offers promising potential for clinical decision support, particularly in managing patients with CUP. INTRODUCTION Accurate tumor-type localization is crucial for determining appropriate treatment strategies, optimizing clinical outcomes, and enabling access to targeted therapies and clinical

FAQs

What is OncoChat and what task does it perform?

OncoChat is an AI model designed to classify tumor types and support tumor-type localization from genomic data, including cancers of unknown primary (CUP).

How was OncoChat developed and evaluated?

It was developed using genomic data from 158,836 tumors sequenced with targeted cancer gene panels, and evaluated against baseline methods using multiple performance metrics. For CUP, it was tested on datasets with confirmed or follow-up validated types and survival associations.

Why does tumor-type classification matter for cancers of unknown primary (CUP)?

Accurate localization improves treatment selection and helps enable targeted therapies and clinical trial access. CUP often leads to less effective empirical chemotherapy because the primary site is unknown, so improved identification can improve clinical outcomes.

Sep 23, 2026

Uploaded by Mary Man