Dimensionality reduction for visualizing single-cell data using UMAP

This paper evaluates uniform manifold approximation and projection (UMAP) as a dimensionality-reduction technique for visualizing single-cell data. Single-cell technologies allow high-resolution dissection of tissue composition but produce large numbers of parameters, and several tools exist to analyze them. UMAP is a recently developed nonlinear method designed for any high-dimensional data, and the authors apply it to biological data using three well-characterized mass cytometry and single-cell RNA sequencing datasets.

Comparing UMAP's performance against five other dimensionality-reduction tools, the authors find that UMAP provides the fastest run times, the highest reproducibility, and the most meaningful organization of cell clusters. On this basis the work highlights UMAP as offering improved visualization and interpretation of single-cell data relative to the other tools tested.