Table 8.
Test data set | |||
---|---|---|---|
Model | GRCh37 | GRCh38 | GRCm38 |
Radial using GRCh37, GRCm38 and first ORF | |||
Sensitivity | 9 8 . 8 6 % | 9 9 . 4 2 % | 98.51% |
Specificity | 97.56% | 9 7 . 6 9 % | 97.54% |
Accuracy | 9 8 . 2 1 % | 9 8 . 5 5 % | 98.02% |
Radial using GRCh37, GRCm38 and longest ORF | |||
Sensitivity | 98.05% | 98.67% | 97.60% |
Specificity | 97.53% | 97.59% | 97.54% |
Accuracy | 97.79% | 98.13% | 97.57% |
Radial using GRCh38, GRCm38 and first ORF | |||
Sensitivity | 91.22% | 99.24% | 9 8 . 6 6 % |
Specificity | 9 8 , 6 5 % | 97.46% | 97.41% |
Accuracy | 94.93% | 98.35% | 9 8 . 0 3 % |
Radial using GRCh38, GRCm38 and longest ORF | |||
Sensitivity | 98.31% | 98.20% | 98.23% |
Specificity | 97.83% | 97.63% | 9 7 . 7 4 % |
Accuracy | 98.07% | 97.91% | 97.98% |
We trained four models with two data sets, GRCh37/GRCm38 and GRCh38/GRCm38, and also compared the selection of two attributes, first and longest ORF relative lengths. The best results for each test data set, GRCh37, GRCh38 and GRCm38, are in bold