Fig. 4: Decomposition of testis gene expression patterns with sparse decomposition of arrays (SDA).
From: Diverse monogenic subforms of human spermatogenic failure
a We applied sparse decomposition analysis (SDA) to testis scRNA-seq data from 12 human donors to identify latent factors (‘components’) representing gene modules. These components are defined by two vectors – one that indicates the loading of each cell on the component, and one that indicates the loading of each gene on the component. b The same scRNA-seq data was summarized using a conventional tSNE analysis, and testicular cell types labels assigned to all cells. c By plotting the cell scores for three representative germ cell components on the tSNE, and as a function of pseudotime, it is apparent that transcription during spermatogenesis can be modeled as series of components overlapping in time, coming on and off gradually on different timescales29. Shown are components with activity that start in spermatogonia (35), spermatocytes (59), and spermatids (92).
