Fig. 3
From: Tensorial blind source separation for improved analysis of multi-omic data
Comparison of multi-way algorithms on a multi-tissue smoking EWAS. a Left panel: Box plot of sensitivity (SE) values for each of the seven methods as applied to the data tensors of dimension 2×152×1062 (two tissues, 152 samples and 1000 randomly selected non-smkDMCs plus 62 smkDMCs) and for 100 different selections of non-smkDMCs. SE(Max) is the maximum sensitivity to capture 62 smkDMCs among all inferred components. Right panel: Heat map of the corresponding one-tailed paired Wilcoxon rank sum test, benchmarking the SE values of each method (y-axis) against each other method (x-axis). b As a, but now for the combined sensitivity (SE(All)) obtained from all enriched components. c,d As a,b, but now for data tensors of dimension 2×152×10 062 and for 100 randomly selected 10 000 non-smkDMCs. EWAS epigenome-wide association study, SE sensitivity, smkDMC smoking-associated differentially methylated CpG, SP specificity