Performance of 17 ESSTs on Detecting Active Sites.

Out of 81,410 residues in the test-sets, 602 residues are annotated as ��ACT_SITE�� by UniProt [17] or CSA [16]. For those active sites, CRESCENDO [8] could either correctly predict (TP) or fail to predict (FN) (see text). Two active sites of ��d7odca1�� (A chain of PDB 7odc), which is a SCOP domain in the test-sets, was discarded as of an internal error; hence, 600 active sites either in the TP or FN. The number of predicted residues is same as the sum of TP and FP for each ESST type. Note that residues only from the first cluster of predicted residues (rank 1) were considered in this analysis. TP: True Positive, FP: False Positive, FN: False Negative, TN: True Negative, SENS: Sensitivity, SPEC: Specificity, COV: Coverage.在测试集的81410个残基中，有602个残基被UniProt[17]或CSA[16]注释为活性位点（ACT_SITE）。针对这些活性位点，CRESCENDO[8]的预测结果可分为正确预测（真阳性，True Positive，TP）与预测失败（假阴性，False Negative，FN）两类（详见正文）。测试集中的SCOP结构域d7odca1（即PDB 7odc的A链）的2个活性位点因内部错误被剔除，因此最终有效活性位点共600个，均属于真阳性或假阴性类别。对于每一种ESST类型，预测残基的总数等于真阳性与假阳性（False Positive，FP）的数量之和。请注意，本分析仅纳入排名第1的首个预测残基簇中的残基。其中，TP：真阳性（True Positive），FP：假阳性（False Positive），FN：假阴性（False Negative），TN：真阴性（True Negative），SENS：灵敏度（Sensitivity），SPEC：特异性（Specificity），COV：覆盖率（Coverage）。