Skip to main content

Table 1 Comparison of performance in terms of FDR, mdFDR, TPR, TPR_hypo and TPR_hyper of four procedures (logistic regression, DSS, z-test and the proposed Bayesian) at different γ values for the first set of simulation studies (without subject effect). Results are averaged over 100 replications of 20,000 CpG loci with two samples under each condition

From: A full Bayesian partition model for identifying hypo- and hyper-methylated loci from single nucleotide resolution sequencing data

   Logistic DSS z-test Bayesian
   regression    
γ=0.01 FDR 0.0043 0.0001 0.0055 0.0091
  mdFDR 0.0048 0.0001 0.0059 0.0109
  TPR 0.2156 0.0837 0.2027 0.2548
  TPR_hypo 0.2162 0.0752 0.2129 0.2596
  TPR_hyper 0.2145 0.0952 0.1889 0.2471
γ=0.05 FDR 0.0158 0.0003 0.0182 0.0184
  mdFDR 0.0159 0.0003 0.0183 0.0188
  TPR 0.3861 0.1672 0.3528 0.4099
  TPR_hypo 0.3921 0.159 0.3694 0.4251
  TPR_hyper 0.3787 0.1772 0.3326 0.3912
γ=0.1 FDR 0.0207 0.0005 0.0244 0.0216
  mdFDR 0.0207 0.0005 0.0244 0.0216
  TPR 0.5398 0.2934 0.4964 0.5775
  TPR_hypo 0.5379 0.2851 0.5011 0.5879
  TPR_hyper 0.542 0.3032 0.4909 0.5652
γ=0.15 FDR 0.0208 0.0005 0.0251 0.0217
  mdFDR 0.0208 0.0005 0.0251 0.0217
  TPR 0.6691 0.4345 0.6307 0.7161
  TPR_hypo 0.6619 0.422 0.627 0.7201
  TPR_hyper 0.6779 0.4497 0.6353 0.7113
γ=0.2 FDR 0.02 0.0006 0.0257 0.0214
  mdFDR 0.02 0.0006 0.0257 0.0214
  TPR 0.7769 0.5751 0.7476 0.8214
  TPR_hypo 0.7708 0.5628 0.7422 0.8253
  TPR_hyper 0.7848 0.591 0.7546 0.8163