About the Paper
GeSubNet infers subtype-specific gene interaction networks by combining patient gene expression profiles with prior gene networks from knowledge databases. A deep generative model separates disease subtypes, a graph neural network learns from the prior gene network, and an inference step joins the two to generate a network for each subtype.
Joint work with Zheng Chen, Xin Liu, Rikuto Kotoge, Peng Chen, Yasuko Matsubara, Yasushi Sakurai and Jimeng Sun.