Skip to main content

Table 1 Differential peak calling on simulated data

From: MMDiff: quantitative testing for shape changes in ChIP-Seq data sets

Profile changes  
  TP FP FN TN eFDR (%) SN (%) SP (%) auROC (%)
DESeq 0 + 0.0 0.7 + 0.8 100 + 0.0 9799.3 + 0.8 NaN 0 + 0.0 100 + 0 50 + 0
DBChIP 0.1 + 0.3 2.4 + 1.7 99.9 + 0.3 9797.6 + 1.7 NaN 0.1 + 0.3 100 + 0 50 + 0
DIME 0 + 0.0 1.4 + 1.0 100 + 0.0 9798.6 + 1.0 NaN 0 + 0 100 + 0 50 + 0
GMD 17.8 + 9.1 5.5 + 2.9 82.2 + 9.1 9794.5 + 2.9 26 + 10 17.8 + 9.1 99.9 + 0 83 + 0
MMDiff 34.6 + 4.1 0.7 + 0.8 65.4 + 4.1 9799.3 + 0.8 2 + 0 34.6 + 4.1 100 + 0 83 + 0
Affinity changes  
  TP FP FN TN eFDR (%) SN (%) SP (%) auROC (%)
DESeq 27.0 + 5.6 0.7 + 0.8 73.0 + 5.6 9799.3 + 0.8 2 + 0 27.0 + 5.6 100 + 0 81 + 0
DBChIP 50.1 + 3.8 2.4 + 1.7 49.9 + 3.8 9797.6 + 1.7 4 + 0 50.1 + 3.8 100 + 0 94 + 0
DIME 45.3 + 4.1 1.4 + 1.0 54.7 + 4.1 9798.6 + 1.0 3 + 0 45.3 + 4.1 100 + 0 95 + 0
GMD 2.1 + 2.1 5.5 + 2.9 97.9 + 2.1 9794.5 + 2.9 73 + 30 2.1 + 2.1 99.9 + 0 60 + 10
MMDiff 2.5 + 1.5 0.7 + 0.8 97.5 + 1.5 9799.3 + 0.8 NaN 2.5 + 1.5 100 + 0 70 + 0
  1. Performance summary of five different methods on ten runs of simulated data sets. In the upper panel unchanged sites and sites with profile changes are considered, in the lower panel unchanged sites and sites with affinity changes. FDR threshold: 0.05; TP: true positives, FP: false positives, FN: false negatives, TN: true negatives, eFDR: empirical FDR (FP/(FP+TP)), SN: sensitivity, SP: specificity, auROC: area under ROC.