17 ESSTs and the Number of Functional Residues Masked from the Alignments.

New ESSTs were based on the structure alignments of SCOP families [19]. ENZ is 221 enzyme-specific SCOP families which contain at least one ACT_SITE annotation of UniProt [17] or hand-curated CSA entry [16]. NOENZ is the opposite of ENZ. NOENZ does not even contain the predicted entries of CSA. ALL is the final alignment source obtained from the filtering process (see Materials and Methods). The masking sources of A, B, C, and D are in Table 1. X is for non-masking and R is for random-masking. R is set as a control to see the significance of removing functional residues from the substitution models. The ESST of Shi et al. (OLD-J) [11] is based on 177 HOMSTRAD families which consist of 706 structures. It masks 2,048 resides which are involved in (1) interaction with heteroatoms and (2) domain�Cdomain interaction. OLD-X and OLD-R is non-masking and random-masking model of J.aNumber of all residues.bNumber of masking residues.c%Mask = number of masking residues/number of all residues*100.