Fig. 1From: scGAL: unmask tumor clonal substructure by jointly analyzing independent single-cell copy number and scRNA-seq dataThe workflow of scGAL. scGAL aims to identify clonal copy number substructure from single-cell copy number data, by borrowing complementary information from independent scRNA-seq data of the same cell line. An autoencoder (AE) is employed to learn the latent representations of scDNA-seq cells, and a generative adversarial network (GAN) is used to mimic the distribution of real scRNA-seq data given the representations. The unsupervised representation learning with combined usage of AE and GAN enables effective reconstruction of the underlying intercorrelations between single-cell copy number and gene expression data, thus making it possible to get more refined clonal copy number substructure. Based on the learned latent representations, a Gaussian mixture model is then used to identify the cell subpopulations. CNP: copy number profileBack to article page