Application of generative AI to heterogenous transcriptome data of AD postmortem brain tissues

초록

In the domains of natural language processing and image processing, generative artificial intelligence (AI) has undergone remarkable growth, exerting a notable impact on our daily lives. However, in other domains, such progress has been less pronounced, warranting further exploration. In our previous studies, we successfully applied generative AI in the study of Alzheimer's disease (AD) in mice. Our objective now is to extend and refine this approach for human AD studies. As widely acknowledged, human data is often inaccessible for experimental purposes, and obtaining brain tissues at specific time points is unfeasible. Our research applying generative AI seeks to overcome these challenges. Our overarching goal is to categorize individuals into typical normal control (NC), typical AD, and on-pathway groups, defining the typical AD progression. Moreover, we seek to generate unobserved intermediate states within this progression using generative AI. Before embarking on a comprehensive application of generative AI, our initial focus was on understanding the diverse genetic types and associated genes in individuals. Analytical methods such as co-clustering and SCENIC were employed as part of this initiative to delineate these genetic types.