modPDZpep: Strucure-based analysis of human PDZ-mediated interaction networks

Home About Tutorial Bound Templates Benchmark Analyze experimentally known PDZ-peptide interactions Contact us


BENCHMARKING RESULTS:
* Resampling was done to balance the positive and negative dataset and average values are recorded.
S.No Criteria of selecting PDZ domains #PDZ domains #Positive interactions #Real negative interactions #Artificial negative interactions Average ROC-AUC Average PR-AUC
1 Domains with real negative data 43 1428 365 - 0.71 0.75
2 Domains with real negative data and scoring ROC-AUC>0.6 34 1179 310 - 0.79 0.81
3 Domains with real and artificial negative data 57 1961 365 1400 0.70 0.71

Other Statistical parameters for 34 domains with real negative data
(Random set of equal positive and negative peptides was used.)
Statistical Parameter Value (Cutoff: -2.11)
Sensitivity (%) 68.62
Specificity (%) 70
False Positive Rate (FPR %) 30
Positive Predictive Value (PPV %) 69.58
Accuracy (%) 69.31
F1 Score 0.69


Individual domain ROC-AUC and PR-AUC
Click to download benchmarking dataset: