MSK Harnesses AI and Real-World Genomic Data to Advance Lung Cancer Care
核心洞察
Memorial Sloan Kettering Cancer Center is leveraging artificial intelligence and large language models to extract clinical data from electronic health records at unprecedented scale, enabling insights across hundreds of thousands of patients.
The MSK-CHORD dataset harmonizes clinicogenomic data for all patients who have undergone tumor genomic profiling, a scale achievable only through AI-assisted curation rather than human chart review.
A machine learning algorithm that predicts cancer (搜索)-associated blood clots (搜索) has received an FDA investigational device exemption under the Software as a Medical Device pathway, a first for MSK.
Memorial Sloan Kettering Cancer Center (MSK) is advancing oncology through the integration of artificial intelligence and large-scale genomic profiling, according to Justin Jee, MD, PhD, a thoracic medical oncologist and physician-scientist who leads MSK's Clinical Data Mining Team. Dr. Jee's work centers on using AI to predict how patients will respond to cancer (搜索) treatments and to improve their quality of life, reflecting a broader institutional effort to translate real-world data into clinical insight.
Dr. Jee's laboratory uses large language models (LLMs) to read electronic health records and extract key information from free-text notes. He notes that this task has traditionally been performed by medical students, residents, and fellows and is "very time consuming." According to Dr. Jee, "LLMs can do these tasks quite well, often better than humans, which can allow us to extract data from hundreds of thousands of patients."
This capability has been applied to all patients at MSK who have undergone tumor genomic profiling, resulting in what the team calls the Clinicogenomic Harmonized, Oncologic Real-world Dataset, or MSK-CHORD. Dr. Jee emphasizes that "the scale of MSK-CHORD would be impossible with human curators alone." The dataset enables improved treatment matching to clinical trials and the discovery of clinically significant relationships, such as how certain genomic subtypes of cancer (搜索) respond to immunotherapy.
Beyond data extraction, Dr. Jee develops AI models deployed directly in the clinic. One example is a machine learning algorithm that predicts which patients will develop blood clots (搜索), which he describes as "an unfortunately common cause of morbidity for patients with cancer (搜索)." After obtaining an investigational device exemption from the FDA under the Software as a Medical Device pathway — described as "a first for MSK" — Dr. Jee now serves as principal investigator of a clinical trial testing whether this AI algorithm can help determine how long a patient should remain on a blood thinner after a cancer-associated blood clot.
MSK's data infrastructure underpins these efforts. To date, over 100,000 patients have had tumors sequenced with MSK technology, either through DNA sequencing with MSK-IMPACT, liquid biopsy profiling with MSK-ACCESS, or RNA sequencing with MSK-TARGET. Clinical data annotations for these patients, extracted from electronic health records using AI and other methods, are now available. All of this data is accessible on cBioPortal, a widely adopted interface developed at MSK, to anyone with an MSK email address.
Dr. Jee also serves as Co-Director of the Thoracic Liquid Biopsy program, a role he inherited from his former mentor, Bob Li. Together with thoracic surgeon James Isbell, Li helped build a bank of thousands of blood samples from patients with lung cancer (搜索) and established collaborations with several companies to test new liquid biopsy technologies.
Reflecting on the broader field, Dr. Jee describes the current era as "a very special time to be a lung cancer (搜索) doctor," citing the emergence of KRAS (搜索) inhibitors, TP53 (搜索)-restoring agents, CAR-T cells, and mRNA vaccines entering the solid tumor space, alongside immunotherapy, antibody drug conjugates, and multimodal therapy with radiation and surgery. He notes that since joining MSK as a fellow in 2019, multiple new FDA-approved treatments for lung cancer have emerged, with measurable impacts on mortality in the overall population. Notably, Dr. Jee observes that "what's happening in the drug development space is even more exciting than what's happening in AI right now."