Skip to main content
Fig. 2 | Genome Medicine

Fig. 2

From: A reference profile-free deconvolution method to infer cancer cell-intrinsic subtypes and tumor-type-specific stromal profiles

Fig. 2

Comparison of cell fractions estimated by different methods. Comparing tumor purity (cancer cell fraction) estimation by different deconvolution methods with the one based on ABSOLUTE (treated as “ground truth”) in terms of Spearman’s correlation coefficients (a) or median absolute deviation (b). The p values above are the difference between DeClust and other methods according to the two-sided paired t test. c Scatter plot of tumor purity estimates by DeClust or ESTIMATE_tr against the one by ABSOLUTE for each cancer type (ESTIMATE_tr transformed ESTIMATE ssGSEA score to fit ABSOLUTE estimates.). d Correlation among the top 3 deconvolution methods (i.e., DeClust, ESTIMATE_ssGSEA, and CIBERSORT_abs) in estimations of immune cell fractions (red), and correlation among the top 3 deconvolution methods (i.e., DeClust, EPIC, and ESTIMATE_ssGSEA) in estimations of stromal cell fractions (gray)

Back to article page