Large scale hierarchical clustering of protein sequences

BMC Bioinformatics

Table 1 Comparison of SYSTERS, TribeMCL and single linkage clustering (SLC) on Pfam and ENZYME data sets (J: Jaccard Coeffcient, SE: Sensitivity, SP: Specificity). The best result in each row is shown in bold face. For the single linkage clustering only the results of the "best" clustering are shown together with the corresponding cutoff E-value. In the case of Sensitivity/Specificity these values were choosen according to the intercept point of the two curves when plotting the values for all possible E-value cutoffs. All clustering procedures were applied to the non-redundant data set and redundant sequences were added to the cluster sets again to compare to the "true" cluster sets: ^a33,963,365 pairwise values of 283,113 non-redundant sequences used for clustering and 442,872 redundant sequences used in comparison; ^b1,582,948 pairwise values of 38,176 non-redundant sequences used for clustering and 84,405 redundant sequences used in comparison.

ISSN: 1471-2105