ZINB-Based Graph Embedding Autoencoder for Single-Cell RNA-Seq Interpretations

Single-cell RNA sequencing (scRNA-seq) provides high-throughput information about the genome-wide gene expression levels at single-cell resolution, bringing a precise understanding on transcriptome of individual cells. Unfortunately, rapidly growing scRNA-seq data and prevalence dropout events pose substantial challenges for cell type annotation. Here, we propose model-based deep graph embedding clustering (scTAG) method, which simultaneously learns cell–cell topology representations identifies clusters based convolutional network. scTAG integrates zero-inflated negative binomial (ZINB) model into adaptive autoencoder to learn low-dimensional latent representation adopts Kullback–Leibler (KL) divergence tasks. By optimizing loss, ZINB reconstruction jointly optimizes cluster label assignment feature learning with topological structures preserved in an end-to-end manner. Extensive experiments 16 RNA-seq datasets from diverse yet representative platforms demonstrate superiority over various state-of-the-art methods.