Skip to main content

Table 1 Prediction results of TFs with good PBM quality by using BayesPI2 energy-independent model and energy-dependent model including dinucleotide interactions

From: Quality versus accuracy: result of a reanalysis of protein-binding microarrays from the DREAM5 challenge by using BayesPI2 including dinucleotide interdependence

  TF family Rank CorrCoef (Ind) Length (Ind) Number (Ind) CorrCoef (Dep) Length (Dep) Number (Dep)
TF_7 bHLH 1 0.78 13 1 0.786 10 1
TF_26 bHLH 2 0.66 10 1 0.67 13 1
TF_56 C2H2 Z F(4) 3 0.76 12 1 0.788 13 1
TF_55 AT hook 4 0.8 8 1 0.8 9 1
TF_17 NR 5 0.76 9 1 0.757 8 1
TF_11 NR 6 0.826 10 1 0.833 10 1
TF_16 Myb/SANT 7 0.808 10 1 0.817 9 1
TF_31 C2H2 ZF (13) 8 0.585 12 1 0.59 11 1
TF_15 Pou + Homeo 9 0.62 11 1 0.655 11 1
TF_45 Myb/SANT 11 0.8 12 1 0.78 11 1
TF_42* Forkhead 12 0.75 12 1 0.805 13 1
TF_64 C2H2 ZF (3) 13 0.75 8 1 0.75 10 1
TF_52 NR 14 0.807 12 1 0.79 10 1
TF_3* Forkhead 16 0.67 10 1 0.724 11 1
TF_27* bZIP 17 0.526 9 1 0.635 11 1
TF_18 Sox 18 0.638 8 1 0.677 8 1
TF_22* T-box 19 0.675 9 1 0.746 12 1
TF_47 Homeo 21 0.726 12 1 0.767 11 1
TF_44 GATA 22 0.633 12 1 0.68 10 1
TF_28 C2H2 ZF (8) 23 0.58 11 1 0.6 11 1
TF_13* Pou + Homeo 24 0.584 9 1 0.675 13 1
TF_5 C2H2 ZF (3) 25 0.59 10 1 0.618 11 1
TF_43 Forkhead 27 0.577 10 1 0.614 12 1
TF_19 Sox 29 0.559 8 1 0.59 11 1
TF_39 C2H2 ZF (3) 30 0.63 11 1 0.668 13 1
TF_51* Pou + Homeo 31 0.615 12 1 0.67 13 1
TF_23 T-box 33 0.64 12 1 0.66 13 1
TF_12* NR 35 0.55 13 1 0.606 13 1
TF_49 NR 34 0.675 10 1 0.676 11 1
TF_53* RFX 39 0.696 12 1 0.77 13 1
TF_14 Myb/SANT 40 0.6 8 1 0.62 9 1
TF_48 NR 43 0.755 12 1 0.78 12 1
TF_38 DM 45 0.67 9 1 0.689 8 1
TF_32 C2H2 ZF (6) 55 0.587 9 1 0.62 8 1
  1. In the table, the 34 TFs were classified by applying fuzzy neuronal gas algorithm on the paired PBM quality-control parameters (i.e. the length of the major and minor axes of the PCA ellipses), where a good agreement between training and testing PBMs indicates good PBM data quality; Rank means TFs are sorted in decreasing order of their final performance score across all tested algorithms in Figure 2 of original publication [1]; CorrCoef , Length, and Number are Pearson correlation between predicted intensities and testing probe intensities, the length of motif, the first or second motif, respectively; (Ind) and (Dep) represent BayesPI2 energy-independent model and energy-dependent model including dinucleotide interaction, respectively; TFs marked by star and bold text indicate that the increase in Pearson correlation coefficient is greater than 0.05 by using BayesPI2 energy-dependent model including dinucleotide interaction energies.