Explainable AI-Based Multi-Omics Analysis of Lung Cancer Health Disparity - PROJECT SUMMARY The proposed research aims to understand and discover the disparity of lung cancer development between African Americans and European Americans, considering race and sex. For example, Among American Males, the highest number of deaths occur from lung cancer, which is more deaths than kidney, pancreas, and prostate cancers combined. African American Males have significantly higher death rates in lung cancer compared to European American Males. Cigarette smoking is considered the strongest risk factor for lung cancer. However, smoking alone cannot explain the disparity in lung cancer development between African American Males and European American Males. The traditional cohort-based genome-wide association studies failed to identify the African American Male-specific genetic locations susceptible to causing lung cancer. These studies are similar to the current standard of medical practice, which largely relies on population-based epidemiological studies in which individuals' genetic and epigenetic variabilities are largely ignored, resulting in population-based conclusions. As a result, similar cancer types respond differently to the same treatment since each tumor has a unique set of mutations. A personalized approach that can identify the patient-specific genetic and epigenetic alterations leading to the disparities in lung cancer incidence and mortality may overcome the issues with population-based studies. Recent advancement in explainable artificial intelligence shows that by applying the local interpretation based on game theory, one can discover the patient-specific significant genes related to a disease condition by analyzing the gene expression profile of a patient. We propose to develop an explainable artificial intelligence- based computational tool to explore multi-omics data in deciphering the lung cancer health disparity by discovering, first, patient-specific disparity information, second, cohort-specific disparity information combining individual disparity information, and third, disparity between two cohorts by comparing cohort-specific disparity information, which will lead to the discovery of cohort-specific risk factors. The proposed tool can be applied to find the cohort-specific risk factors by creating and comparing cohorts based on race and sex, such as African American Males vs. European American Males, African American Females vs. European American Females, etc. Thus, the developed tool will help solve the disparity puzzle that exists in lung cancer incidence and mortality in different races and ethnicities. The proposed work is important since it will identify the heterogeneity of lung cancer at a personalized level, which may help precision medicine's rational design. This improved precision medicine design will help reduce lung cancer health disparities between different ethnic and racial communities and maximize the therapeutic benefit in all communities.
