Skip to main content

Table 4 The prediction performance of Scansite, NetPhosK, KinasePhos and GPS for four well-studied PKs of PKA, CK2, ATM and S6K.

From: PPSP: prediction of PK-specific phosphorylation site with Bayesian decision theory

  PK Group PKA CK2 ATM S6K
Predictor   Sn(%) Sp(%) MCC Sn(%) Sp(%) MCC Sn(%) Sp(%) MCC Sn(%) Sp(%) MCC
PPSP Defaulta 90.11 91.7 0.3841 83.21 90.01 0.3596 93.02 94.06 0.4627 92.85 97.97 0.6618
ScanSite Highb 21.98 99.96 0.4450 10.95 99.86 0.2655 18.6 99.8 0.3443 N/A N/A N/A
  Medium 44.51 99.39 0.5084 27.01 99.11 0.3342 25.58 98.89 0.2756 N/A N/A N/A
  Low 47.8 98.29 0.4041 54.02 96.34 0.3684 51.16 94.89 0.2739 N/A N/A N/A
NetPhosK Default 79.12 90.65 0.3165 82.48 89.43 0.3464 86.01 98.51 0.6786 82.35 97.14 0.5404
KinasePhos 90% (Sp)d 90.72 91.3 0.3783 72.53 91.58 0.3384 88.37 87.8 0.3137 N/A N/A N/A
  95% (Sp) 89.18 94.62 0.4595 64.58 94.93 0.3806 88.37 92.14 0.3893 N/A N/A N/A
  100% (Sp) 76.8 98.47 0.6154 54.86 98.66 0.5222 86.05 96.89 0.5497 N/A N/A N/A
GPS Default 88.88 90.57 0.3564 82.99 87.59 0.3210 90.86 89.55 0.3498 94.9 91.34 0.3964
  1. a. The default parameters are employed for PPSP, NetPhosK and GPS.
  2. b. ScanSite 2.0 has three thresholds for prediction, including high, medium and low stringencies.
  3. c. N/A – not available.
  4. d. KinasePhos has paid attention to prediction specificity with three cut-off values, as 90%, 95% and 100%.