Comparative analyses of seven algorithms for copy number variant identification from single nucleotide polymorphism arrays

Abstract

Determination of copy number variants (CNVs) inferred in genome wide single nucleotide polymorphism arrays has shown increasing utility in genetic variant disease associations. Several CNV detection methods are available, but differences in CNV call thresholds and characteristics exist. We evaluated the relative performance of seven methods: circular binary segmentation, CNVFinder, cnvPartition, gain and loss of DNA, Nexus algorithms, PennCNV and QuantiSNP. Tested data included real and simulated Illumina HumHap 550 data from the Singapore cohort study of the risk factors for Myopia (SCORM) and simulated data from Affymetrix 6.0 and platform-independent distributions. The normalized singleton ratio (NSR) is proposed as a metric for parameter optimization before enacting full analysis. We used 10 SCORM samples for optimizing parameter settings for each method and then evaluated method performance at optimal parameters using 100 SCORM samples. The statistical power, false positive rates, and receiver operating characteristic (ROC) curve residuals were evaluated by simulation studies. Optimal parameters, as determined by NSR and ROC curve residuals, were consistent across datasets. QuantiSNP outperformed other methods based on ROC curve residuals over most datasets. Nexus Rank and SNPRank have low specificity and high power. Nexus Rank calls oversized CNVs. PennCNV detects one of the fewest numbers of CNVs.

Deletion Variants of RABGAP1L, 10q21.3, and C4 Are Associated With the Risk of Systemic Lupus Erythematosus in Korean Women

“…We used the PennCNV algorithm, which is one of the most commonly used programs for CNV identification from SNP array data (16). Although PennCNV is known to have some limitations in detecting small-sized CNVs and its performance was rated intermediate in a study by Dellinger et al (which evaluated the performances of various methods), its relatively low false-positive rates support the reliability of its call rates (17).…”

Copy Number Variation Accuracy in Genome-Wide Association Studies

“…First, a recent study suggested that disease-related CNVs detected from GWAS data are well tagged by SNPs, and, therefore, CNVs do not add further information [10] . Second, there is evidence that different methods for identifying CNVs from GWAS data report different results, even when applied to the same array data [11] .…”

“…CNVPartition uses a likelihood-based algorithm, PennCNV implements a hidden Markov model, and QuantiSNP uses an objective Bayes hidden-Markov model. A detailed comparison of these different algorithms can be found in the study by Dellinger et al [11] . These 3 programs have often helped to find putative disease-related CNVs [18][19][20][21][22] .…”

The genome-wide landscape of copy number variations in the MUSGEN study provides evidence for a founder effect in the isolated Finnish population

“…Most importantly, differences in the array architecture, choice of algorithms, population differences and phenotypes affect our understanding of CNVs. [48][49][50][51] Being aware of all such discrepancies, we chose to abide by widely adopted methods to make our data comparable across studies. However, an increase in the sample size and a higher marker density in this study would fine-tune the estimates of population frequencies and their correlation with the studied phenotypes.…”