Table 4 ROC AUC and PR AUC scores of a multi-omic based logistic regression and CNN trained with real patient MRIs, cGAN predicted MRIs and a combination of real and predicted MRIs for TP53, PIK3CA and CDH1 with testing set containing 20 percent of total
Gene (# of mutated patients/total patients) | ROC AUC | PR AUC |
|---|---|---|
TP53 multi-omic logistic regression (235/690) | 0.9400 | 0.9009 |
TP53 with real MRI (21/50) | 1.0000 | 1.0000 |
TP53 with cGAN MRI (235/690) | 0.9477 | 0.8926 |
TP53 with cGAN + real MRI (256/740) PIK3CA multi-omic logistic regression(247/690) PIK3CA with real MRI (11/50) PIK3CA with cGAN MRI (247/690) PIK3CA with cGAN + real MRI (258/740) | 0.9508 0.7209 0.8333 0.7407 0.7515 | 0.9301 0.5413 0.8110 0.6360 0.7184 |
CDH1 multi-omic logistic regression (112/690) | 0.8068 | 0.4342 |
CDH1 with real MRI (11/50) | 0.9167 | 0.9083 |
CDH1 with cGAN MRI (112/690) | 0.8105 | 0.4861 |
CDH1 with cGAN + real MRI (123/740) | 0.8136 | 0.5007 |