Skip to main content
. Author manuscript; available in PMC: 2020 Sep 1.
Published in final edited form as: Hum Mutat. 2019 Jun 24;40(9):1314–1320. doi: 10.1002/humu.23825

Table 1.

Evaluation metrics for all submissions and for the baseline method. The table is broken up by submissions that used the known warfarin confounding, those that did not, and the baseline method. Within each group scores are sorted by AUC. Accuracy, sensitivity, specificity, and F1 are calculated using a cutoff of 0.5 for all predictions.

Description Submission Approach AUC Accuracy Sensitivity Specificity F1
Did not use warfarin in prediction Group 5a Unsupervised 0.65 0.51 0.26 0.95 0.40
Group 1b Scoring 0.60 0.60 0.59 0.59 0.65
Group 5c Unsupervised 0.59 0.63 0.70 0.49 0.70
Group 3 Scoring 0.59 0.34 0.23 0.54 0.31
Group 1a Scoring 0.57 0.47 0.30 0.76 0.42
Group 5d Unsupervised 0.53 0.59 0.73 0.32 0.69
Group 2 Scoring 0.49 0.41 0.12 0.92 0.21
Group 7a Unsupervised 0.48 0.41 0.21 0.76 0.31
Group 5b Unsupervised 0.47 0.53 0.65 0.30 0.64
Used warfarin in prediction Group 4 Supervised 0.76 0.70 0.71 0.65 0.75
Group 6a Scoring 0.65 0.72 0.85 0.46 0.79
Group 6d Unsupervised 0.61 0.64 0.70 0.51 0.71
Group 6c Unsupervised 0.44 0.47 0.53 0.35 0.56
Group 6b Unsupervised 0.43 0.47 0.56 0.30 0.57
Soria et al. Baseline Genetic risk score 0.71 0.67 0.68 0.65 0.73