A new artificial intelligence model called COMPASS, developed by Harvard Medical School researchers and their colleagues, improves prediction of which patients are most likely to respond to immune checkpoint inhibitors (ICIs), according to a July 3 report. The model uses data from patients treated in the past and outperformed existing approaches by 8.5 percent, making predictions based on tumor gene activity and providing a rationale for its output.
If validated in future clinical trials, COMPASS could help personalize cancer treatment, streamline trial enrollment for new therapies, and identify potential drug targets for further research. The results were published July 3 in Nature Medicine.
Immune checkpoint inhibitors were first approved by the U.S. Food and Drug Administration in 2011. These drugs target proteins such as PD-L1, PD-1, and CTLA-4 that can shield cancer cells from immune attack. By disrupting these proteins' interactions with T cells or tumor cells, ICIs allow the immune system to recognize and destroy cancer cells.
Although some patients experience significant benefits from ICIs—such as former U.S. president Jimmy Carter surviving nine years after a stage IV melanoma diagnosis—clinical trials have shown only 10 percent to 40 percent of patients respond positively depending on their cancer type. Nonresponders may face serious side effects without effective treatment.
Marinka Zitnik said, "Understanding who will respond to ICIs is not a minor knowledge gap. It is one of the central unsolved problems in oncology." Zitnik's team designed COMPASS using concept bottleneck transformer architecture so that its predictions are interpretable rather than black-box outputs.
The researchers trained COMPASS with data from over 10,000 tumors across multiple cancer types using information from the Cancer Genome Atlas public database and fine-tuned it with results from clinical trials involving different ICI regimens. In testing scenarios where individual clinical trials were withheld during training, COMPASS consistently outperformed previous models across various conditions including different cancers and sequencing platforms.
Zitnik said if these findings hold up in prospective clinical trials, "COMPASS could find use in cancer clinics as a decision aid to help doctors decide which individuals would benefit most from ICIs." She added that interpretable results could also generate new hypotheses about how the immune system fights cancer.
