Criteria for optimization of consensus prediction. Comparisons of different consensus prediction criteria for the subcellular localization (A), transmembrane β-barrel (B) and lipoprotein (C) predictors tested on a modified PA training dataset using four statistical parameters: accuracy, recall/sensitivity, specificity and Mathews Correlation Coefficient (MCC). Four criteria for subcellular localization predictors, four for transmembrane β-barrel predictors, and two for lipoprotein predictors were considered. The statistical parameters evaluated on each criterion are shown on the x-axis. A score of one (y-axis) indicates the highest performance and zero the lowest.