Extended Data Table 4 Comparison of in silico prediction models with BRCA2 DBD MAVE functional results

From: Functional evaluation and clinical classification of BRCA2 variants

  1. The sensitivity was calculated using MAVE functional results as ground truth and comparing the MAVE P_Strong evidence thresholds with predicted pathogenic/likely pathogenic class (i.e. C65 from AGVGD, likely pathogenic from AlphaMissense, pathogenic from BayesDel, and P Strong/Moderate/Supporting for BayesDel with ClinGen thresholds), or comparing the MAVE B_Strong evidence thresholds with predicted benign/likely benign class (i.e. C0 from AGVGD, likely benign from AlphaMissense, benign from BayesDel, and B Moderate/Supporting for BayesDel with ClinGen thresholds). Formula for the sensitivity calculation is shown below:
  2. (MAVE P_strong and predicted P/LP) / (all MAVE P_strong) or (MAVE B_strong and predicted B/LB) / (all MAVE B_strong)
  3. The specificity was calculated using MAVE functional results as ground truth and comparing the MAVE non-P_Strong evidence thresholds with predicted non-pathogenic/non-likely pathogenic class, or comparing the MAVE non-B_Strong evidence thresholds with predicted non-benign/non-likely benign class. Formula for the specificity calculation is shown below:
  4. (MAVE non-P_strong and predicted non-P/LP) / (all MAVE non-P_strong) or (MAVE non-B_strong and predicted non-B/LB) / (all MAVE non-B_strong)
Back to article page