Fig. 3
From: Assessing characteristics of RNA amplification methods for single cell RNA sequencing
Single-cell RNA-sequencing precision. a Pairwise correlations for all samples. Upper triangle: Pearson correlation. Lower triangle: Kendall correlation. Zeros treated as missing values. Each row and column is an individual sample. Experimental group is indicated by color bars at edge of plot. b Relationship between standard deviation (st. dev.) and mean characterized by least squares regression (see Methods ). All estimated coefficients were highly significant (coefficient t-test p < 10−16). c–f Enrichment of biophysical traits in experimentally precise (low) and variable (high) genes with respect to background genes (see Methods ). Error bars indicate 95% CI. “*” indicates significant difference (p < 0.05). Numbers at bottom indicate sample size (number of genes). (C) Median difference in gene length estimated by Hodges-Lehman statistic. Significance: Wilcoxon rank sum two-way test. (D) Relative risk of containing an internal A-hexamer. Significance: Fisher’s exact test. (E) As C for % GC content. (F) As C for strength of local secondary structure. g–j PCA projection of dilution data on PC 1 and 2. Plots were centered so that bulk UHR or HBR was positioned at the origin. Points represent individual dilution replicates. Colored ovals represent bivariate normal 95% confidence ellipses. % Sd. explained by a PC is indicated in axis label. See Methods . (G) HBR 10 pg. (H) UHR 10 pg. (I) HBR 100 pg. (J) UHR 100 pg. k–n As G–J, but using only abundantly expressed genes (see main text). Axis scales differ from G–J, with axes in equivalent to the purple-boxed region in G–J. (K) HBR 10 pg. (L) UHR 10 pg. (M) HBR 100 pg. (N) UHR 100 pg. Abbreviations: Standard error (SE); confidence interval (CI); kilobases (kb); kilocalories (kcal); mole (mol); principal components analysis (PCA); principal component (PC); standard deviation (sd)