Large language model trained on clinical oncology data predicts cancer progression
Abstract Subspecialty knowledge barriers have limited the adoption of large language models (LLMs) in oncology. We introduce Woollie, an open-source, oncology-specific LLM trained on real-world data from Memorial Sloan Kettering Cancer Center (MSK) across lung, breast, prostate, pancreatic, and colo...
Saved in:
| Main Authors: | , , , , , , , , , , , , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2025-07-01
|
| Series: | npj Digital Medicine |
| Online Access: | https://doi.org/10.1038/s41746-025-01780-2 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1849389305530679296 |
|---|---|
| author | Menglei Zhu Hui Lin Jue Jiang Abbas J. Jinia Justin Jee Karl Pichotta Michele Waters Doori Rose Nikolaus Schultz Sulov Chalise Lohit Valleru Olivier Morin Jean Moran Joseph O. Deasy Shirin Pilai Chelsea Nichols Gregory Riely Lior Z. Braunstein Anyi Li |
| author_facet | Menglei Zhu Hui Lin Jue Jiang Abbas J. Jinia Justin Jee Karl Pichotta Michele Waters Doori Rose Nikolaus Schultz Sulov Chalise Lohit Valleru Olivier Morin Jean Moran Joseph O. Deasy Shirin Pilai Chelsea Nichols Gregory Riely Lior Z. Braunstein Anyi Li |
| author_sort | Menglei Zhu |
| collection | DOAJ |
| description | Abstract Subspecialty knowledge barriers have limited the adoption of large language models (LLMs) in oncology. We introduce Woollie, an open-source, oncology-specific LLM trained on real-world data from Memorial Sloan Kettering Cancer Center (MSK) across lung, breast, prostate, pancreatic, and colorectal cancers, with external validation using University of California, San Francisco (UCSF) data. Woollie surpasses ChatGPT in medical benchmarks and excels in eight non-medical benchmarks. Analyzing 39,319 radiology impression notes from 4002 patients, it achieved an overall area under the receiver operating characteristic curve (AUROC) of 0.97 for cancer progression prediction on MSK data, including a notable 0.98 AUROC for pancreatic cancer. On UCSF data, it achieved an overall AUROC of 0.88, excelling in lung cancer detection with an AUROC of 0.95. As the first oncology specific LLM validated across institutions, Woollie demonstrates high accuracy and consistency across cancer types, underscoring its potential to enhance cancer progression analysis. |
| format | Article |
| id | doaj-art-0e81809abeba4f61bc8491f0afda3d43 |
| institution | Kabale University |
| issn | 2398-6352 |
| language | English |
| publishDate | 2025-07-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | npj Digital Medicine |
| spelling | doaj-art-0e81809abeba4f61bc8491f0afda3d432025-08-20T03:42:00ZengNature Portfolionpj Digital Medicine2398-63522025-07-018111510.1038/s41746-025-01780-2Large language model trained on clinical oncology data predicts cancer progressionMenglei Zhu0Hui Lin1Jue Jiang2Abbas J. Jinia3Justin Jee4Karl Pichotta5Michele Waters6Doori Rose7Nikolaus Schultz8Sulov Chalise9Lohit Valleru10Olivier Morin11Jean Moran12Joseph O. Deasy13Shirin Pilai14Chelsea Nichols15Gregory Riely16Lior Z. Braunstein17Anyi Li18Memorial Sloan Kettering Cancer CenterUniversity of California San FranciscoMemorial Sloan Kettering Cancer CenterMemorial Sloan Kettering Cancer CenterMemorial Sloan Kettering Cancer CenterMemorial Sloan Kettering Cancer CenterMemorial Sloan Kettering Cancer CenterMemorial Sloan Kettering Cancer CenterMemorial Sloan Kettering Cancer CenterMemorial Sloan Kettering Cancer CenterMemorial Sloan Kettering Cancer CenterUniversity of California San FranciscoMemorial Sloan Kettering Cancer CenterMemorial Sloan Kettering Cancer CenterMemorial Sloan Kettering Cancer CenterMemorial Sloan Kettering Cancer CenterMemorial Sloan Kettering Cancer CenterMemorial Sloan Kettering Cancer CenterMemorial Sloan Kettering Cancer CenterAbstract Subspecialty knowledge barriers have limited the adoption of large language models (LLMs) in oncology. We introduce Woollie, an open-source, oncology-specific LLM trained on real-world data from Memorial Sloan Kettering Cancer Center (MSK) across lung, breast, prostate, pancreatic, and colorectal cancers, with external validation using University of California, San Francisco (UCSF) data. Woollie surpasses ChatGPT in medical benchmarks and excels in eight non-medical benchmarks. Analyzing 39,319 radiology impression notes from 4002 patients, it achieved an overall area under the receiver operating characteristic curve (AUROC) of 0.97 for cancer progression prediction on MSK data, including a notable 0.98 AUROC for pancreatic cancer. On UCSF data, it achieved an overall AUROC of 0.88, excelling in lung cancer detection with an AUROC of 0.95. As the first oncology specific LLM validated across institutions, Woollie demonstrates high accuracy and consistency across cancer types, underscoring its potential to enhance cancer progression analysis.https://doi.org/10.1038/s41746-025-01780-2 |
| spellingShingle | Menglei Zhu Hui Lin Jue Jiang Abbas J. Jinia Justin Jee Karl Pichotta Michele Waters Doori Rose Nikolaus Schultz Sulov Chalise Lohit Valleru Olivier Morin Jean Moran Joseph O. Deasy Shirin Pilai Chelsea Nichols Gregory Riely Lior Z. Braunstein Anyi Li Large language model trained on clinical oncology data predicts cancer progression npj Digital Medicine |
| title | Large language model trained on clinical oncology data predicts cancer progression |
| title_full | Large language model trained on clinical oncology data predicts cancer progression |
| title_fullStr | Large language model trained on clinical oncology data predicts cancer progression |
| title_full_unstemmed | Large language model trained on clinical oncology data predicts cancer progression |
| title_short | Large language model trained on clinical oncology data predicts cancer progression |
| title_sort | large language model trained on clinical oncology data predicts cancer progression |
| url | https://doi.org/10.1038/s41746-025-01780-2 |
| work_keys_str_mv | AT mengleizhu largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT huilin largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT juejiang largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT abbasjjinia largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT justinjee largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT karlpichotta largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT michelewaters largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT doorirose largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT nikolausschultz largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT sulovchalise largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT lohitvalleru largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT oliviermorin largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT jeanmoran largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT josephodeasy largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT shirinpilai largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT chelseanichols largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT gregoryriely largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT liorzbraunstein largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression AT anyili largelanguagemodeltrainedonclinicaloncologydatapredictscancerprogression |