Imperfect gold standard gene sets yield inaccurate evaluation of causal gene identification methods

Causal gene discovery methods are often evaluated using gold-standard (GS) sets of causal genes. However, GS gene sets are always incomplete, leading to mis-estimation of sensitivity, specificity, AUC. Labeling biases in GS gene sets can also lead to inaccurate ordering of discovery methods. We argue that evaluation of these methods should rely on statistical techniques like those used for variant discovery, rather than on comparison with GS gene sets.