scAEspy: a unifying tool based on autoencoders for the analysis of single-cell RNA sequencing data

Abstract

Autoencoders (AEs) have been effectively used to capture the non-linearities among gene interactions of single-cell RNA sequencing (scRNA-Seq) data. However, their integration with the common scRNA-Seq bioinformatics pipelines still poses a challenge. Here, we introduce scAEspy, a unifying tool that embodies five of the most advanced AEs and different loss functions, including two novel AEs that we developed. scAEspy allows the integration of data generated using different scRNA-Seq platforms. We benchmarked scAEspy against principal component analysis (PCA) on five public datasets, showing that our new AEs outperform the existing solutions, achieving more than 20% increase of the Rand Index in the identification of cell clusters.