netNMF-sc: leveraging gene–gene interactions for imputation and dimensionality reduction in single-cell expression analysis

(Downloading may take up to 30 seconds. If the slide opens in your browser, select File -> Save As to save it.)

Click on image to view larger version.

Figure 5.
Figure 5.

Comparison of differential expression of marker genes and gene–gene correlations in untransformed data from Buettner et al. (2015) and data imputed using netNMF-sc, NMF, scImpute, and MAGIC. (A) Overlap between differentially expressed genes and periodic genes (log P-values from Fisher's exact test). (B) Expression of the G1/S phase marker gene Exo1 in cells labeled as G1/S (blue) and cells labeled as G2/M (green) in data imputed by each method. In netNMF-sc inputed data, Exo1 is overexpressed in G1/S cells compared to G2/M cells (P ≤ 6.7 × 10−12), as expected. In contrast, in data imputed by MAGIC, Exo1 is underexpressed in G1/S cells compared to G2/M cells (P ≤ 2.2 × 10−16). Exo1 shows no difference in expression in untransformed and scImpute data. (C) Distribution of R2 correlation coefficients between pairs of periodic genes in the cell cycle data. (D) Scatter plot of expression of two G1/S phase genes, Dtl and Exo1, across cells. These genes are positively correlated in data imputed by netNMF-sc (P ≤ 2.2 × 10−16), negatively correlated in data imputed by MAGIC (P ≤ 2.2 × 10−16), and uncorrelated in other methods.

This Article

  1. Genome Res. 30: 195-204

Preprint Server