Fig. 3.

Estimation accuracy of cellular compositions using LM22 + C1 scGEPs. a Scatter plots of the estimated and true cell proportions for the 20 simulated bulk tumor samples. Each dot represents one sample and r denotes the Pearson’s correlation coefficient. b Estimation accuracy of LM22 + C1 scGEPs and CIBERSORT microarray GEPs. Estimation accuracy is measured as the Pearson’s correlation coefficient between true cell proportions and the estimated proportions. The value of Pearson’s correlation coefficients is coded by both area and color of the pie charts for CIBERSORT microarray GEPs (top) and LM22 + C1 scGEPs (bottom). Larger pie slices and darker blue represent larger Pearson’s correlation coefficients thus higher accuracy. The missing cell types in CIBERSORT microarray GEPs are denoted by dashes. T cell composition is calculated taking sum of the four T cell subtypes