Pre-capture multiplexing provides additional power to detect copy number variation in exome sequencing

BMC Bioinformatics

Table 3 mcCNV (MC)/ExomeDepth (ED) calls for ‘WGS’ pool (used as prediction) versus the ERDS/cnvpytor calls from matched genome sequencing (used as truth)

			MCC	TPR	FDR	PPV	BalAcc
DUP + DEL	Full	MC	0.18	0.34	0.90	0.10	0.67
	Full	ED	0.26	0.36	0.81	0.19	0.68
	Sub	MC	0.49	0.34	0.31	0.69	0.67
	Sub	ED	0.48	0.38	0.38	0.62	0.69
DUP	Full	MC	0.40	0.24	0.33	0.67	0.62
	Full	ED	0.35	0.24	0.50	0.50	0.62
	Sub	MC	0.40	0.25	0.33	0.67	0.62
	Sub	ED	0.38	0.27	0.45	0.55	0.63
DEL	Full	MC	0.18	0.64	0.95	0.05	0.82
	Full	ED	0.22	0.56	0.91	0.09	0.78
	Sub	MC	0.68	0.66	0.29	0.71	0.83
	Sub	ED	0.54	0.55	0.47	0.53	0.78

Calls are subdivided by duplications (DUP) and deletions (DEL). ‘Full’ gives performance across the full pool; ‘Sub’ gives the performance excluding the poorly correlated samples NCG_00790 and NCG_00851 (gray rows). ‘MCC’ is Matthew’s correlation coefficient, ‘TPR’ is true positive rate/sensitivity, ‘FDR’ is false discovery rate, ‘PPV’ is positive predictive value, ‘BalAcc’ is balanced accuracy. Exons with any overlap of the repetitive and low-complexity regions, as defined in the Trost et al. manuscript, omitted from analysis

ISSN: 1471-2105