Researchers at Mount Sinai and Memorial Sloan Kettering believe they’ve made a breakthrough in a new form of cancer treatment that has taken the oncology field by storm.
The scientists are among several groups trying to use AI to make it easier to determine how an individual will react to a new therapy, known as immune checkpoint inhibitors, that uses the body’s own immune system to target malignant cells. Data published by the group this week shows their AI-based software is more effective and less invasive than existing methods of assessing how a patient will react to the therapy, a necessary step before beginning treatment.
In the mid-1990s, scientists discovered that human T cells – sentries in the blood that help the body fight infection – could be deputized to attack cancer by disabling the immune system “checkpoints” that regulate their flow. The first immune checkpoint inhibitor therapy was approved by the Food and Drug Administration in 2011 and since then, the cancer treatment has blossomed into a $4.8 billion industry.
While the therapy itself can be less toxic than chemotherapy, it does come with its own side effects, like fatigue and inflammation, and the test to determine whether it will work can be invasive, time-consuming and costly. Current methods of assessing patients’ eligibility for the therapy require a tumor biopsy and complex genomic testing.
The partners have developed a predictive computer software known as Scorpio, which they say could determine eligibility by tracing cell counts, enzymes, proteins and other particles in standard blood samples coupled with clinical data in a patient’s electronic medical records, rather than requiring a biopsy.
“In theory, the data should already be there for this patient, you wouldn’t even need to do anything additional to get it,” said Dr. Luc Morris, a surgeon and research lab director at Memorial Sloan Kettering and one of the leaders of the project.
Memorial Sloan Kettering, the world’s oldest and largest private cancer center, has a vast reserve of data for an AI model to study. Overall, the tests have involved patient data of more than 10,000 individuals, with 21 types of cancer ranging from common diagnoses like breast cancer to less common malignancies, like sarcoma, to ensure that the model was exposed to a wide spectrum of possibilities. The model was first tested on data from 2,000 patients from Memorial Sloan Kettering, then again on another 2,100, and finally on information from Mount Sinai patients to ensure the program was effective on different datasets.
A new study led by the group and published this week shows the program outperformed existing FDA-approved assessments both in terms of predicting the clinical benefits and overall survival of the patient. The research was backed by the National Institutes of Health and the Department of Defense with seed funding from Mount Sinai.
“Routine blood tests represent a very important tool in modern medicine, so all this information is easily available in the electronic medical records in every hospital, in every clinic,” said Dr. Diego Chowell, an assistant professor at Mount Sinai’s Icahn School of Medicine and a co-leader of the project.
The collaboration takes advantage of the complementary skills of Morris and Chowell’s teams. Memorial Sloan Kettering collected and curated the data to train the model, while Mount Sinai’s computational researchers did much of the machine learning work, Morris said.
The researchers hope the program will one day be licensed for use on a global scale. The group will seek FDA approval over the coming months and partnerships with other hospitals to fine-tune the algorithm, Chowell said.
