Table I.

Relative performance differences of the two clustering methods

The table provides a comparison of the most important strengths and weaknesses (Performance Criteria) when the two approaches BCL and HCL are used with the clustering parameters of this study. The reported differences are based on benchmark comparisons with curated families. Altered settings can change the performance differences significantly. The “++” stands for better performance than “+” in this table.

Performance Criteria	BCL	HCL
High sensitivitya	+	++
Low false-positive rate under high-sensitivity settings	+	++
Elimination of clusters with inconsistent domain composition	+	++
Rejection of clusters with unconnected domains	+	++
Control over length of similarity region relative to sequences	++	+
Not limited by domain knowledge	++	+

^aReferences: Eddy (1996); Altschul et al. (1997).