Supervised dimensionality reduction for exploration of single-cell data by Hybrid Subset Selection - Linear Discriminant Analysis

Single-cell technologies generate large, high-dimensional datasets encompassing a diversity of omics. Dimensionality reduction enables visualization of data by representing cells in two-dimensional plots that capture the structure and heterogeneity of the original dataset. Visualizations contribute to human understanding of data and are useful for guiding both quantitative and qualitative analysis of cellular relationships. Existing algorithms are typically unsupervised, utilizing only measured features to generate manifolds, disregarding known biological labels such as cell type or experimental timepoint. Here, we repurpose the classification algorithm, linear discriminant analysis (LDA), for supervised dimensionality reduction of single-cell data. LDA identifies linear combinations of predictors that optimally separate a priori classes, enabling users to tailor visualizations to separate specific aspects of cellular heterogeneity. We implement feature selection by hybrid subset selection (HSS) and demonstrate that this flexible, computationally-efficient approach generates non-stochastic, interpretable axes amenable to diverse biological processes, such as differentiation over time and cell cycle. We benchmark HSS-LDA against several popular dimensionality reduction algorithms and illustrate its utility and versatility for exploration of single-cell mass cytometry, transcriptomics and chromatin accessibility data.