scAAnet

scAAnet is a software tool for data analysis for single-cell RNA sequencing (scRNA-seq). It is an autoencoder-based method that performs non-linear archetypal analysis to identify gene expression programs (GEPs) and infer their relative activity across individual cells. Unlike traditional cluster analysis, scAAnet can reveal a continuous spectrum of cellular states and shared GEPs across different cell types.

Key features of scAAnet include:

1. A count distribution-based loss term to handle the sparsity and overdispersion of raw scRNA-seq count data.

2. An archetypal constraint was added to the loss function to improve the interpretability of the identified GEPs.

3. As demonstrated through simulations, Superior performance compared to existing archetypal analysis methods.

4. Ability to extract biologically meaningful GEPs from publicly available scRNA-seq datasets, including pancreatic islet, lung idiopathic pulmonary fibrosis, and prefrontal cortex data.