Single‑cell RNA sequencing (scRNA‑Seq) provides detailed insight into gene expression at the individual cell level, revealing hidden cell diversity. However, scRNA‑Seq data pose challenges due to high‑dimensionality and sparsity. High‑dimensionality stems from analysing numerous cells and genes, while sparsity arises from zero counts in gene expression data, known as dropout events. This necessitates robust data processing methods of the scRNA‑Seq gene counts, for meaningful interpretation. Dimensionality reduction techniques, such as principal component analysis, transform gene count data into lower‑dimensional spaces retaining biological information, aiding in downstream analyses, while dimensionality reduction‑based visualisation methods, such as t‑distributed stochastic neighbour embedding, and uniform manifold approximation and projection are used for cell or gene clustering. Deep learning techniques, such as variational autoencoders and generative adversarial networks compress data and generate synthetic gene expression profiles, augmenting datasets and improving utility in biomedical research. In recent years, the interest for scRNA‑Seq dimensionality reduction has markedly increased, not only leading to the development of a multitude of methods, but also to the integration of these approaches into scRNA‑Seq data processing pipelines. The present review aimed to list and explain, in layman's terms, the current popular dimensionality reduction methods, as well as include advancements and software package implementations of them.

Related Articles