Comparing segmentations by applying randomization techniques

BMC Bioinformatics

Table 2 Conditional entropies and their significances for isochore segmentations with 114 segments

P	H (T\|P)	p _ℓ	p _k	H (P\|T)	p _ℓ	p _k
L _gc	0.740	0	0	0.683	0	0
L ₂	0.688	0	0	0.771	0	0
L ₁	0.750	0	0	0.778	0	0
L _A	0.937	0.0352	0.0389	0.876	0.0010	0.0090
L _C	0.797	0	0	0.684	0	0
L _G	0.836	0.0002	0.0004	0.821	0.0001	0.0005
L _T	0.984	0.1758	0.1748	0.840	0.0002	0.0010

Conditional entropies and entropy distances for segmentations on isochore structure, for the ground truth segmentation from [20] with k = 114. L_f: least-squares segmentation with features f; f ∈ {gc, 2, 1, A, C, G, T} indicate frequencies of G+C, 2-letter words, 1-letter words, and frequency of A, C, G, or T, respectively, p_ℓ: the fraction of segmentations from $C_{N, k, ℓ}$ with a smaller value of the conditional entropy in 10,000 randomizations; p_k: the fraction of segmentations from $C_{N, k}$ with a smaller value of the conditional entropy in 10,000 randomizations.

ISSN: 1471-2105