CARAT: A novel method for allelic detection of DNA copy number changes using high density oligonucleotide arrays

BMC Bioinformatics

Table 4 Comparison among CARAT, dCHIP and CNAG using QPCR results.

Method	Stage	Sensitivity	Specificity
CARAT	Single Point	0.956 (0.932)	0.444 (0.4)
	100 kb Smoothing	0.957 (0.898)	0.556 (0.6)
	Tree Partitioning	0.957 (0.932)	0.778 (0.8)
	Tree Partitioning (p-value)	0.957 (0.881)	1 (1)
dCHIP	Single Point	0.783 (0.780)	0.333 (0.4)
	10 Points Median	0.864 (0.913)	0.7 (0.667)
	HMM	0.978 (0.898)	0.778 (0.8)
CNAG	Single Point	0.609	1
	10 points Mean	0.522	1
	HMM	0.717	1

Note: SNPs that have no copy number alterations relative to the diploid state (true negative SNPs) are defined by a QPCR derived copy number between 1.5 and 2.5. SNPs that have copy number alterations (true positive SNPs) are defined by a QPCR derived copy number less than 1.5 or greater than 2.5. There are in total 10 true negatives among all 69 SNPs. For the 55 SNPs that are common to CNAG, there are 9 true negatives. The calculations of sensitivity and specificity compare QPCR values to algorithm outputs in which algorithm-defined true negatives are the SNPs with an estimated copy number between 1.5 and 2.5 and algorithm-defined true positives are the SNPs with an estimated copy number less than 1.5 or greater than 2.5. In the CARAT tree-partitioning (p-value) comparison, the algorithm defined negative SNPs are those with p-value > 0.005 and algorithm defined positive SNPs are those with p-value < 0.005. Numerical values without parentheses are derived from comparisons using only the 55 SNPs that are in common with CNAG while the numbers inside parentheses are derived from comparisons using all 69 SNPs.

ISSN: 1471-2105