Table 1 Comparison of Results for SMERFS and MINER. ROC_0.1 scores for optimised parameters of SMERFS and MINER (window size, W, gap threshold, C _g and scoring scheme), as tested in the blind data subset, separate from the training data. In the 'Combined' site type, protein and small molecule binding residues are pooled to form a single standard set. In the central scoring scheme, scores produced by SMERFS or MINER are applied to the central position of the window only, in 'max', all alignment positions take the score of the window covering them with the maximum score. For the 0–0.1 false positive rate range here, a random prediction would on average produce an area of 0.005.

From: The contrasting properties of conservation and correlated phylogeny in protein functional residue prediction