Skip to main content

Table 3 Performance comparison of various prediction algorithms for ER binding using the independent dataset shown in Table 2

From: Multiplatform genome-wide identification and modeling of functional human estrogen receptor binding sites

Prediction algorithm Sensitivity Specificity Harmonic mean Fisher's exact test Pvalue
Consensus ERE with ≤2 mismatches 94.74% 30% 45.57% 0.104838
Consensus ERE with ≤3 mismatches 94.74% 0% 0.00% 1
Dragon ERE finder v2.0 68.42% 40% 50.49% 0.477589
TFAC 8.1 (min FP) 31.57% 100% 47.99% 0.057117
TFAC 8.1 (min FN) 78.94% 40% 53.10% 0.255439
h-ERE (stringent) 42.10% 90% 57.37% 0.084693
h-ERE (medium) 68.42% 70% 69.20% 0.056272
h-ERE (relaxed) 73.68% 70% 71.79% 0.03043
h-ERE (loose) 84.21% 70% 76.45% 0.006199
  1. h-ERE outperformed the other algorithms. ERE, estrogen response element.