Table 3.

Performance with the validation dataset of the best in silico tools previously selected from the results at discovery stage.

	Sensitivity	Specificity	Accuracy	MCC	Positive predictive Value	Negative Predictive Value	False Negative Rate	False Positive Rate	False Discovery Rate	False Omission Rate
Donor
HSF
All variants	96.045	90.909	95.238	0.831	98.266	81.081	3.955	9.091	1.734	18.919
Without invariable dinucleotides	94.643	90.909	93.793	0.830	97.248	83.333	5.357	9.091	2.752	16.667
HSF+SSF-like
All variants	99.435	93.939	98.571	0.946	98.876	96.875	0.565	6.061	1.124	3.125
Without invariable dinucleotides	99.107	93.939	97.931	0.941	98.230	96.875	0.893	6.061	1.770	3.125
HSF+SSF-like+MES
All variants	99.435	93.939	98.571	0.946	98.876	96.875	0.565	6.061	1.124	3.125
Without invariable dinucleotides	99.107	93.939	97.931	0.941	98.230	96.875	0.893	6.061	1.770	3.125
Acceptor
MES and SSF-like sequential
All variants	91.579	95.122	92.647	0.837	97.753	82.979	8.421	4.878	2.247	17.021
Without invariable dinucleotides	71.429	95.000	85.294	0.699	90.909	82.609	28.571	5.000	9.091	17.391
SSF-like
All variants	92.632	92.683	92.647	0.832	96.703	84.444	7.368	7.317	3.297	15.556
Without invariable dinucleotides	75.000	92.500	85.294	0.695	87.500	84.091	25.000	7.500	12.500	15.909

The best performance scores are highlighted in bold. The atypical BRCA2 exon 17 native donor site (GC) was not estimated by HSF nor MES, and we have considered it as a failed prediction of the two tools for variants affecting this exon regardless of the in vitro splicing effect of the variant. False Discovery Rate represents the rate of false positives of the total of variants positively predicted and False Omission Rate represents the rate of false negatives of the total negative predicted variants.