神経変性疾患研究におけるマルチオミクス統合解析

抄録

In recent years, rapid developments in omic technologies have enabled the quantitative measurement of multimodal data such as the genome, transcriptome, proteome, and metabolome. In the field of neurodegenerative diseases, several large–scale longitudinal cohort studies have collected not only clinical and pathological information but also multimodal omic data from serum, cerebrospinal fluid, and postmortem brain tissues. These datasets are publicly available to researchers worldwide through platforms such as the AMP–AD Knowledge Portal, allowing for a wide range of integrative analyses. However, the integration of multimodal data remains challenging due to differences in data types, measurement scales, and noise structures. Thus, many multi–omic integrative analytical techniques have been proposed for biomarker discovery, prognosis prediction, and elucidation of molecular disease mechanisms. Integrative approaches can be broadly categorized into “knowledge-based approach” and “data-driven approach”. The former integrates different types of omic data by referencing prior knowledge accumulated in databases. On the other hand, the latter method employs statistical and machine–learning algorithms to extract latent structures or cross–modal relationships directly from data. In our recent study, we constructed a “metabolic trans-omic network” for Alzheimer disease by integrating transcriptomic, proteomic, and metabolomic data from the ROS/MAP study. This network analysis provided a systems–wide view on dysregulated energy metabolism in Alzheimer disease. Overall, integrative multi–omic analysis represents a powerful framework for investigating neurodegenerative diseases. Nonetheless, current approaches still face several limitations, including insufficient model interpretability and challenges in ensuring reproducibility and robustness. Thus, further improvements in integration techniques are anticipated.