Skip to main content

Table 1 Average AUC values for simulation data with replicates

From: TCC: an R package for comparing tag count data with robust normalization strategies

DE method edgeR edgeR edgeR DESeq DESeq DESeq baySeq baySeq baySeq
P G1 50% 70% 90% 50% 70% 90% 50% 70% 90%
(a) TMM          
PDEG = 5% 90.30 90.30 90.18 89.09 89.08 88.92 89.43 89.42 89.22
PDEG = 15% 90.30 90.12 89.63 89.15 88.91 88.26 89.54 89.28 88.66
PDEG = 25% 90.40 89.89 88.26 89.27 88.62 86.82 89.63 89.07 87.25
(b) DEGES/TbT          
PDEG = 5% 90.30 90.31 90.28 89.09 89.12 89.07 89.43 89.47 89.41
PDEG = 15% 90.30 90.25 90.26 89.15 89.07 89.01 89.54 89.48 89.49
PDEG = 25% 90.41 90.27 89.89 89.27 89.07 88.61 89.63 89.56 89.22
(c) DEGES/edgeR          
PDEG = 5% 90.29 90.32 90.28 89.09 89.13 89.08 89.43 89.48 89.41
PDEG = 15% 90.30 90.25 90.26 89.15 89.07 89.02 89.54 89.48 89.50
PDEG = 25% 90.40 90.28 89.91* 89.27 89.08 88.64 89.63 89.56 89.24
(d) iDEGES/edgeR          
PDEG = 5% 90.29 90.32 90.28 89.09 89.13 89.08 89.43 89.48 89.42
PDEG = 15% 90.30 90.25 90.29* 89.15 89.08 89.06 89.54 89.49 89.54
PDEG = 25% 90.40 90.30* 90.04* 89.27 89.11 88.80 89.63 89.60 89.42
(e) iDEGES/TDT          
PDEG = 5% 90.30 90.32 90.28 89.09 89.12 89.09 89.43 89.48 89.42
PDEG = 15% 90.30 90.23 90.20 89.15 89.05 88.93 89.54 89.45 89.40
PDEG = 25% 90.40 90.25 89.82 89.27 89.04 88.53 89.63 89.52 89.13
(f) iDEGES/DESeq          
PDEG = 5% 90.30 90.31 90.28 89.09 89.13 89.09 89.43 89.48 89.43
PDEG = 15% 90.30 90.24 90.21 89.15 89.05 88.95 89.54 89.45 89.42
PDEG = 25% 90.40 90.26 89.86 89.27 89.05 88.58 89.63 89.53 89.18
  1. Average AUC values (%) of 100 trials for each simulation condition are shown. Simulation data contain a total of 10,000 genes: PDEG% of genes is for DEGs, PG1% of PDEG in G1 is higher than in G2, and each group has three biological replicates (i.e., G1_rep1, G1_rep2, G1_rep3, G2_rep1, G2_rep2, and G2_rep3). A total of nine conditions (three PDEG values × three PG1 values) are shown. The highest AUC value for each condition is in bold. AUC values with asterisks indicate significant improvements (p-value < 0.01, paired t-test) compared with DEGES/TbT. We used a bootstrap resampling size of 10,000 in baySeq when performing the normalization (i.e., baySeq at step 2 in the DEGES/TbT) and 2,000 when performing the DEG identification after normalization (i.e., baySeq in the XXX-baySeq combination).