Fig. 9

Evaluation of the impact of surrogate variable analysis (SVA) on (bulk) RNAseq differential expression analysis using pipeComp. RUVr and RUVs refer to the two RUVSeq functions by the same name; svaseq refers to the function of the same name in sva, while vstsva refers to the original sva method applied no variance-stabilized data. a True positive rates (TPR) and false discovery rates (FDR) of the SVA-based methods (using a single surrogate variable), averaged across DEA methods. b TPR and FDR of the different SVA-based methods (averaged across the different DEA methods) using increasing numbers of variables. In a, b, the x-axis is square root transformed to improve readability, and three dots for each combination of methods refer to the nominal FDR thresholds of 0.01, 0.05, and 0.1. c Correlation of estimated logFC with expected ones, as well as TPR and FDR (at nominal FDR threshold of 0.05) of the different methods with increasing numbers of variables. As in Fig. 4c, the color mapping tracks the number of (matrix-wise) median absolute deviations from the (column-wise) median, while the printed numbers represent the raw metric values. be and leek refer to the two methods implemented in the sva package for estimating the number of surrogate variables