Skip to main content

Table 7 Comparison of three existing tools used to identify exonic SAVs with MutPred Splice

From: MutPred Splice: machine learning-based prediction of exonic variants that disrupt splicing

Method ANNOVAR Human splicing finder Skippy MutPred splice
Splicing focus Splice site disruption All exonic and intronic ESE/ESS disruption and cryptic splice site All exonic
Prediction output Binary label Multiple output scores Multiple output scores Probabilistic, with additional hypothesis of splicing mechanism disrupted
TP 41 65 68 (61) 121
FP 4 33 15 7
TN 79 50 68 76
FN 140 (0) 116 113 (57) 60
FPR% 4.8 39.8 18.1 8.4
Sensitivity (%) 22.7 (100.0) 35.9 37.6 (51.7) 66.9
Specificity (%) 95.2 60.2 81.9 91.6
Accuracy (%) 58.9 (97.6) 48.1 59.7 (66.8) 79.2
MCC 0.22 (0.93) -0.04 0.19 (0.34) 0.54
  1. Evaluation was based on 264 exonic variants (181 positive, 83 negative). Performance metrics are given for guidance only as not all tools may be directly comparable (due to different applications or limitations). Performance scores in parentheses reflect adjusted performance based upon the evaluation of only specific categories of splicing mutation (for example, splice site disruption) relevant to the respective tool. For methods that output multiple scores for a variant (HSF and Skippy), performance metrics may differ depending upon the features and thresholds applied. TP, true positives; FP, false positives; TN, true negatives; FN, false negatives; FPR, false positive rate; MCC, Matthews correlation coefficient.