Table 1.

Number of clusters

	0.99	0.97	0.95
DNACLUST exact	233879	73726	28241
DNACLUST inexact	240125	76391	28661
UCLUST exact	144339	48418	20039
UCLUST inexact	253108	71361	26685
CD-HIT	245851	100280	55208

The number of clusters produced by DNACLUST, UCLUST and CD-HIT at various identity/similarity thresholds, on the twins dataset. Since each tool uses slightly different distance measures, the number of clusters can not be directly compared between different tools. (Namely the identity measure used by UCLUST and CD-HIT underestimates the distance between two sequences, as computed by the similarity measure used by DNACLUST). Instead we compare the change in the number of clusters when switching between the exact and inexact modes of each tool - a smaller change indicating better performance.