Help for Area Under ROC curve (AUC) interpretation
Reliability of resultant top 200 candidates depends on predictive power of AraNet for the given query gene set as measured by AUC score from ROC analysis. In original publication of AraNet (Nature Biotechnology 2010), we estimated the proportion of genes newly discovered to be associated with the query gene set as a function of AUC score. It concluded that we may expect about 4 newly discovered gene among the top 200 for the query gene set with AUC > 0.6, 5 for AUC > 0.7, 6 for AUC > 0.85, 7 for AUC > 0.9, whereas 0 newly discovered gene is expected by random expectation (See supplementary figure 12, Nature Biotechnology 28:149).
