Classes and continua of hippocampal CA1 inhibitory neurons revealed by single-cell transcriptomics
Fig 6
Latent factor analysis reveals a common mode of continuous variation that is consistent across cell types and correlates with axon target location.
(A) Latent factor analysis assigns a single number to each cell via a search for the factor values that best predict expression of multiple genes. Mean factor values within each cluster differ systematically in a way that correlates with the identified cell class’s axon target location. Each point represents a cell, with the x-coordinate showing latent factor value and y-coordinate showing cluster, sorted by mean latent factor value. (B) Continuous gradient of latent factor values across continent 8 (top; symbol size denotes latent factor value). Largest values are found in western Slc17a8-expressing neurons identified with soma-targeting Cck basket cells; smallest values are found in eastern Calb1-expressing neurons identified with dendrite-targeting Cck cells. (C) Correlation of latent factor values with expression of six example genes. Symbols as in (A); blue, continent 2 (basket/bistratified), green, continent 8 (Cck/Cxcl14). (D) Correlations of genes with the latent factor are preserved across cell classes. X-axis: correlation of gene with latent factor in continent 2; y-axis, correlation in continent 8; Spearman’s ρ = 0.58, p < 10−100. Inset, Spearman ρ values for all pairs of continents, p < 10−100 in each case. AU, arbitrary units.