Benchmarking interpretability of deep learning for predictive genomics: Recall, precision, and variability of feature attribution
Table 4
Composite scores measured as the geometric mean across recall, precision and ensemble consistency to provide a single quantitative measure of interpretation performance.