[4Xin2] Poster session 2

[4Xin2-16] Genes Feature Extraction Method in Gene Regulatory Networks Based on Distance between Graphs

〇Sho Oshima¹, Yuji Okamoto², Yohei Harada², Mayumi Kamada², Yasushi Okuno^2,3 (1.Faculty of Medicine, Kyoto University, 2.Graduate School of Medicine, Kyoto University, 3.Riken Center for Computational Science)

Keywords:Gene Regulatory Networks, Distance between Graphs, Graph Neural Networks, Personalized Medicine

In the pursuit of personalized medicine, there is an ongoing effort to design cancer treatment drugs based on genetic mutations. Due to the vast number of potential target genes, there is a shift towards drug design based on gene regulatory networks. A key challenge in current cancer drug design is identifying central genes within these networks, known as "Master Regulators." However, there are no clear criteria for identifying Master Regulators, and previous methods have not fully utilized the structural information of gene regulatory networks. Here we show a new method for estimating Master Regulators in the gene regulatory networks specific to individual patients. The proposed method uses distances between graphs constructed through graph neural networks, enabling the estimation of Master Regulators utilizing the structural information of gene regulatory networks. Furthermore, the utility and validity of this method are demonstrated through correlation with survival time analysis and cancer disease association. This approach is expected to facilitate more effective cancer drug design and contribute to the development of optimal treatment strategies for individual patients.