Fig. 6: Exploring the PBMC expression programs, derived by qNMF and competing algorithms. | Communications Biology

Fig. 6: Exploring the PBMC expression programs, derived by qNMF and competing algorithms.

From: q-Diffusion leverages the full dimensionality of gene coexpression in single-cell transcriptomics

Fig. 6

a DEMGs between control and stimulated conditions are more unique to individual PBMC subtypes after q-diffused regularization is imposed on NMF. This effect is evidenced by the greater number of decreases (blue) in Jaccard similarity than increases (red) between the cell-type pairs of DEMGs. Decreases account for 75% and 89% of the off-diagonal entries, respectively. In the diagonals, we list the actual numbers of q-diffused DEMGs. b Total number of GEPs with at least one statistically enriched biological pathway. Competing NMF methods are shown for different amounts of GEPs, in powers of two. We analyzed qNMF, regularized NMF with a Seurat-style neighborhood graph (kNMF), typical NMF, and the recent Projective NMF (PNMF).

Back to article page