Table 2.
Results of the human case study
| Test data set | ||
|---|---|---|
| Model | GRCh37 | GRCh38 |
| Radial using GRCh37 and first ORF | ||
| Sensitivity | 98.95% | 99.43% |
| Specificity | 97.41% | 97.23% |
| Accuracy | 98.18% | 98.33% |
| Radial using GRCh37 and longest ORF | ||
| Sensitivity | 98.09% | 98.73% |
| Specificity | 97.50% | 97.55% |
| Accuracy | 97.80% | 98.14% |
| Quadratic using GRCh37 and first ORF | ||
| Sensitivity | 98.15% | 98.83% |
| Specificity | 96.60% | 96.41% |
| Accuracy | 97.38% | 97.62% |
| Quadratic using GRCh37 and longest ORF | ||
| Sensitivity | 94.79% | 95.54% |
| Specificity | 97.23% | 97.19% |
| Accuracy | 96,01% | 96.36% |
| Radial using GRCh38 and first ORF | ||
| Sensitivity | 89.86% | 97.54% |
| Specificity | 98.64% | 99.26% |
| Accuracy | 94.25% | 98.40% |
| Radial using GRCh38 and longest ORF | ||
| Sensitivity | 98.37% | 97.63% |
| Specificity | 97.76% | 97.58% |
| Accuracy | 98.06% | 97.61% |
| Quadratic using GRCh38 and first ORF | ||
| Sensitivity | 80.43% | 98.66% |
| Specificity | 98.84% | 96.78% |
| Accuracy | 89.63% | 97.72% |
| Quadratic using GRCh38 and longest ORF | ||
| Sensitivity | 94.77% | 95.08% |
| Specificity | 97.66% | 97.50% |
| Accuracy | 96.21% | 96.29% |
We trained 8 models with two data sets, GRCh37 and GRCh38, to select the first, or the longest, ORF relative lengths (the length of the corresponding ORF divided by the length of the transcript). The better results for each data set are in bold