Abstract 3632: Cytokine profiles associated with ICI myocarditis using machine learning approaches identifies novel cytokines and implicated pathways

IntroductionImmune checkpoint inhibitors (ICIs) are effective for a growing number of cancer indications. ICI-mediated T cell activation can lead to immune related adverse effects, including ICI myocarditis, which has up to a 50% mortality. To date, only a few cytokines have been reported to be associated with ICI myocarditis. We evaluated an expansive repertoire of cytokines associated with ICI myocarditis and the pathways they regulate.

MethodsA total of 173 cardio-oncology patients were enrolled in the biomarker study, including 55 who were on ICI. Blood samples of patients who were on ICI for cytokine profiling were sampled when patients presented with symptoms concerning for ICI myocarditis. 71 different cytokines were evaluated and analyzed using Point Biserial correlation analyses and machine learning (XGboost) and explainable artificial intelligence (SHAP) to identify cytokines associated with ICI myocarditis. Analyses were performed to identify pathways associated with ICI myocarditis.

Results28 cytokines were associated with ICI myocarditis and machine learning revealed top features associated with ICI myocarditis in the entire cohort with IL33 being the top feature, Accuracy of 0.895, AUC of 0.902, F1 score of 0.73. SHAP was also used to identify features associated with ICI myocarditis and found IL10, CXCL9, CXCL13, CCL3, were positively associated with ICI myocarditis while CCL22, IL33, TNFSF10, CCL8, and CCL24 were negatively associated with ICI myocarditis. 90% of cytokines identified in the correlation model were also identified in SHAP and XG Boost. Top KEGG and GO pathways associated with ICI myocarditis identified by XGBoost and SHAP features include the cytosolic DNA sensing pathway, response to influenza A, IL17, PI3K-Akt, JAK-STAT and lipid/atherosclerosis pathways

ConclusionsIdentifying pathways associated with ICI myocarditis could provide insights into optimization of immunosuppression strategies

Citation Format: Rachel Jaber Chehayeb, Dat Hong, Nathan W. Chen, Carlos Matute Martinez, Ritujith Jayakrishnan, Ana Ferrigno Guajardo, Derrick Lin, Yunju Im, Stephanie Halene, Jennifer VanOudenhove, John Hwa, Alokkumar Jha, Jennifer M. Kwan. Cytokine profiles associated with ICI myocarditis using machine learning approaches identifies novel cytokines and implicated pathways [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 3632.