Detecting clusters of transcription factors based on a nonhomogeneous poisson process model

BMC Bioinformatics

Table 2 Performance of NLK clustering and four other methods using simulated data in Scenarios 1\(\sim\)3

Methods	AMCR\(^*\)			PPC\(^*\)
	Scenario 1	Scenario 2	Scenario 3	Scenario 1	Scenario 2	Scenario 3
NLK clustering	0.0044	0.0446	0.0798	96.8%	51.1%	0%
Window-based K-means\(^\dag\)	0.0294	0.1006	0.1067	68.9%	21.8%	0%
Window-based Hclust\(^\dag\)	0.0380	0.1222	0.1160	54.2%	11.4%	0%
K-function-based Hclust\(^\ddag\)	0.0763	0.2022	0.1533	43.1%	0.70%	0%
Co-localization-vector-based Hclust\(^\ddag\)	0.0220	0.1184	0.1084	68.2%	11.7%	0%

\(^*\) AMCR: average misclassification rate; PPC: proportion of perfect classification. The boldface number shows the best result across different clustering methods.
\(^\dag\): For these methods, the AMCR and PPC are calculated under the optimal window width such that the AMCR is minimized.
\(^\ddag\): For these methods, the multiple TF binding loci are defined by peaks within a prespecified distance threshold, and the AMCR and PPC are calculated under the optimal distance threshold such that the AMCR is minimized

ISSN: 1471-2105