Presentation Information
[SS06-09]ASURAT: Functional annotation-driven unsupervised clustering of single-cell transcriptomes
*Keita Iida1 (1. The University of Osaka (Japan))
Keywords:
ASURAT,Single-cell omics,Cancer,Semiology
Our bodies contain various types of cells, and life is sustained by the production of numerous molecules through gene expression. It is also true that cells can survive precisely because life is maintained. These mutual relationships extend across multiple layers (molecule, cell, tissue, etc.), forming a common regulatory foundation in living organisms. Therefore, elucidating the nature of this regulation is crucial for understanding both the maintenance of biological functions and their breakdown. However, the information obtained from biological systems is multimodal and high-dimensional, and it also includes many unknown interactions. Consequently, it is difficult to fully uncover the regulatory mechanisms of biological systems using theoretical approaches alone. To address this challenge, we have been developing an approach that integrates mathematical sciences and data science. Here, I will introduce the software ASURAT (Iida, Bioconductor, 2022), which enables multifaceted classification of cells based on single-cell omics data from various biological functional perspectives. A major bottleneck in conventional single-cell omics analysis has been the extensive trial and error and literature review required for cell classification and the biological interpretation of computational results. In contrast, ASURAT mathematically formalizes the concept of "sign," allowing for simultaneous cell classification and interpretation. As an application of ASURAT, I will present findings from spatial transcriptomics of human pancreatic cancer, where we identified functional abnormalities related to cancer malignancy (Iida et al., Bioinformatics, 2022). If time permits, I will also discuss future perspectives, incorporating our latest research findings.