Reconstructing cancer genomes from paired-end sequencing data

BMC Bioinformatics

Table 2 Statistical tests for variant edges

Reciproc al vs. N on Reciprocal Variant Edges
Dataset	VariantType	R (all)	$\bar{R}$ (all)	R (non-triv)	$\bar{R}$ (non-triv)	NR	$\bar{N R}$	p-Val
OV1	T	179	41	75	13	9	58	< 1E-15
OV1	I	46	20	16	12	2	29	3.46E-5
OV1	TO	210	46	70	16	9	38	2.79E-12
OV2	T	77	51	41	23	12	49	5.17E-7
OV2	I	21	15	9	5	10	21	0.057
OV2	TO	96	64	46	18	15	44	2.63E-7
OV3	T	61	13	19	3	6	30	2.111E-7
OV3	I	19	13	5	5	2	13	0.075
OV3	TO	58	26	22	8	7	28	1.92E-5
OV4	T	74	16	40	6	12	35	1.54E-9
OV4	I	10	0	2	0	3	12	0.073
OV4	TO	48	22	22	10	12	26	0.0036
OV5	T	93	19	29	7	8	37	2.30E-8
OV5	I	12	8	2	0	6	13	0.13
OV5	TO	82	26	22	8	7	34	2.29E-6

Results of Fisher's exact test showing that non-trivial reciprocal edges are more likely to be used (assigned a multiplicity μ > 0) in the interval-adjacency graph than non-reciprocal variant edges when a minimum of 5 discordant pairs is required to add a variant edge to the graph. Variant edges are classified as Inversion (I), Translocation (T), and Robertsonian Translocation (TO). Each variant edge is also classified as either reciprocal or not and by whether it is used (μ > 0) or not used (μ = 0). We report the number of edges of the following types: used reciprocal edges (R(all)), non used reciprocal edges ( $\bar{R}$ (all), used reciprocal non-trivial (R(non-triv)), not used reciprocal non-trivial ( $\bar{R}$ (non-triv)), used non-reciprocal (NR), and not used non-reciprocal ( $\bar{N R}$ ).

ISSN: 1471-2105