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

Dellinger AE, Saw SM, Goh LK, Seielstad M, Young TL, Li YJ - Nucleic Acids Res. (2010)

Bottom Line: Nexus Rank and SNPRank have low specificity and high power.Nexus Rank calls oversized CNVs.PennCNV detects one of the fewest numbers of CNVs.

View Article: PubMed Central - PubMed

Affiliation: Center for Human Genetics, Duke University Medical Center, Durham, NC 27710, USA.

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.

Show MeSH

Related in: MedlinePlus

Boxplots of ROC curve residual from simulated data for each method. (a) ROC residuals from simulation 1. (b) ROC residuals from simulation 2.
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC2875020&req=5

Figure 3: Boxplots of ROC curve residual from simulated data for each method. (a) ROC residuals from simulation 1. (b) ROC residuals from simulation 2.

Mentions: Parameters other than those displayed in Supplementary Tables S4 and S6 were tested. Parameters displayed reflect ROC optimal parameters for simulation 1, NSR optimized parameters or parameters with measurable differences from the optimal parameter. Methods ranked from first to last using CNVs were as follows: QuantiSNP and Nexus SNPRank, PennCNV (average ROC residual = 0.707), PennCNV (0.703), CBS (0.675), GLAD (0.631), Nexus Rank (0.598) and CNVFinder (0.061) (Figure 4c; Supplementary Table S4). The interquartile ranges (IQRs) of the highest ranking methods: CBS, Nexus SNPRank, PennCNV and QuantiSNP have the highest ROC residuals and are overlapping (Figure 3a). These methods also displayed comparable sensitivity and specificity rates. Power and false positive rates differed between the CNV and CNV SNP levels, and so ROC values and rankings also differed (Figure 4; Supplementary Table S6).Figure 3.


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

Dellinger AE, Saw SM, Goh LK, Seielstad M, Young TL, Li YJ - Nucleic Acids Res. (2010)

Boxplots of ROC curve residual from simulated data for each method. (a) ROC residuals from simulation 1. (b) ROC residuals from simulation 2.
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC2875020&req=5

Figure 3: Boxplots of ROC curve residual from simulated data for each method. (a) ROC residuals from simulation 1. (b) ROC residuals from simulation 2.
Mentions: Parameters other than those displayed in Supplementary Tables S4 and S6 were tested. Parameters displayed reflect ROC optimal parameters for simulation 1, NSR optimized parameters or parameters with measurable differences from the optimal parameter. Methods ranked from first to last using CNVs were as follows: QuantiSNP and Nexus SNPRank, PennCNV (average ROC residual = 0.707), PennCNV (0.703), CBS (0.675), GLAD (0.631), Nexus Rank (0.598) and CNVFinder (0.061) (Figure 4c; Supplementary Table S4). The interquartile ranges (IQRs) of the highest ranking methods: CBS, Nexus SNPRank, PennCNV and QuantiSNP have the highest ROC residuals and are overlapping (Figure 3a). These methods also displayed comparable sensitivity and specificity rates. Power and false positive rates differed between the CNV and CNV SNP levels, and so ROC values and rankings also differed (Figure 4; Supplementary Table S6).Figure 3.

Bottom Line: Nexus Rank and SNPRank have low specificity and high power.Nexus Rank calls oversized CNVs.PennCNV detects one of the fewest numbers of CNVs.

View Article: PubMed Central - PubMed

Affiliation: Center for Human Genetics, Duke University Medical Center, Durham, NC 27710, USA.

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.

Show MeSH
Related in: MedlinePlus