Abstract We propose to combine mass spectrometry and machine learning to address racial disparities in Alzheimer's disease (AD) research. While African Americans have two to three times the incidence rates of AD than people of European Ancestry, they are underrepresented in past and current AD research studies. The lack of representation leads to a weaker understanding of disease progression and poorer diagnostics for this population. The overarching premise, that advances in diagnosing and treating Alzheimer's disease for all Americans will be stymied unless differences among racial groups are considered, is supported by the literature and demonstrated with preliminary data in the application. The aims of this proposal are, therefore, designed to specifically address the need for a better understanding of Alzheimer's disease in minority groups and to increase the diversity of patients that could be correctly diagnosed with serum-based biomarker panels for AD. Successful completion of the proposed research would accomplish several objectives: We would identify optimal feature selection methods for `omics data; generate the best possible plasma biomarker panel for AD in African Americans; and provide a systematic investigation of the extent to which existing MS- `omics studies' accuracy of predicting AD are modulated by the participants' race.