Table 3

Sensitivities and positive predictive values (PPVs) of HexDiff and other algorithms. A known CRM was considered recovered if a predicted CRM overlapped it by at least 50 bp. The PPVs in this table are italicized because they are estimates of the true PPVs. Without complete knowledge of all CRMs that are present in the 16 sequences, it is possible that some of the predicted CRMs that are labeled as false positives are actually true positives.

Algorithm
CRMs Recovered
Num CRMs
Sensitivity TP/(TP + FN)
True Positives
CRMs Predicted
PPV TP/(TP + FP)

HexDiff
36
52
69.23%
35
104
33.65%
Ahab
23
52
44.23%
20
35
57.14%
Cluster Buster
31
52
59.62%
23
88
26.14%
MSCAN
34
52
65.38%
42
226
18.58%
MCAST
43
52
82.69%
53
499
10.62%
LWF
27
52
51.92%
48
433
11.09%

Chan and Kibler BMC Bioinformatics 2005 6:262   doi:10.1186/1471-2105-6-262