GAN Learning Methods for Bulk RNA-Seq Data and Their Interpretive Application in the Context of Disease Progression.

Abstract

A generative adversarial network (GAN) is a generative model that consists of two adversarial networks, a discriminator and a generator, usually in the form of neural networks. One of the useful things about applying GANs is that they can synthesize two states to produce an intermediate output that implies a semantic feature. When applied to omics data that determine phenotypes of a disease, GANs can be used to associate these intermediate outputs with the progression of the disease. In this chapter, to realize the above idea, we will introduce the application of GAN methods to bulk RNA-seq data, which cover data preprocessing, training, and latent interpolation between different phenotypes describing disease progression.