Figure 3

A new method for essential amino acid prediction. We compare the new method (based on the two sets of amino acids Union and Intersection) with method 1 (based on closeness centrality) and the Conseq server. This method is better than the others: indeed, if we predict fewer than 35% of the amino acids, then the curve (9) is, over most of that range, the highest one; if we predict more than 60% of the amino acids, the curve (10) is, over most of that range, the highest. The calculations are made as an average over the five networks representing HIV-1 protease, TEM1 beta-lactamase, T4 lysozyme, Barnase and the bacteriophage f1 gene V protein.