Technical considerations for genotyping multi-allelic copy number variation (CNV), in regions of segmental duplication

Abstract

Background: Intrachromosomal segmental duplications provide the substrate for non-allelic homologous recombination, facilitating extensive copy number variation in the human genome. Many multi-copy gene families are embedded within genomic regions with high levels of sequence identity (>95%) and therefore pose considerable analytical challenges. In some cases, the complexity involved in analyzing such regions is largely underestimated. Rapid, cost effective analysis of multi-copy gene regions have typically implemented quantitative approaches, however quantitative data are not an absolute means of certainty. Therefore any technique prone to degrees of measurement error can produce ambiguous results that may lead to spurious associations with complex disease.

Results: In this study we have focused on testing the accuracy and reproducibility of quantitative analysis techniques. With reference to the C-C Chemokine Ligand-3-like-1 (CCL3L1) gene, we performed analysis using real-time Quantitative PCR (QPCR), Multiplex Ligation-dependent Probe Amplification (MLPA) and Paralogue Ratio Test (PRT). After controlling for potential outside variables on assay performance, including DNA concentration, quality, preparation and storage conditions, we find that real-time QPCR produces data that does not cluster tightly around copy number integer values, with variation substantially greater than that of the MLPA or PRT systems. We find that the method of rounding real-time QPCR measurements can potentially lead to mis-scoring of copy number genotypes and suggest caution should be exercised in interpreting QPCR data.

Conclusions: We conclude that real-time QPCR is inherently prone to measurement error, even under conditions that would seem favorable for association studies. Our results indicate that potential variability in the physicochemical properties of the DNA samples cannot solely explain the poor performance exhibited by the real-time QPCR systems. We recommend that more robust approaches such as PRT or MLPA should be used to genotype multi-allelic copy number variation in disease association studies and suggest several approaches which can be implemented to ensure the quality of the copy number typing using quantitative methods.

Figure 1

Schematic representation of the CCL3L1 gene located on chromosome 17q12. Overlapping segmental duplications are indicated by the black and grey arrows and the probe and primer positions for the PRT, real-time QPCR and MLPA assays are indicated by labels.

Figure 2

The distribution of unrounded CCL3L1 copy number measurements plotted on a histogram from the PRT and MLPA exon 1 assays. A. The MLPA assay demonstrates clear clustering, with peaks centered on integer values and gaps between each of the clusters ranging from 0–4 copies. B. The PRT assay demonstrates a distribution of 0–5 copies, with peaks clearly centered on integer values.

Figure 3

The distributions of unrounded CCL3L1 copy number measurements plotted on a histogram from real-time QPCR and MLPA exon 3 assays. A. The real-time QPCR assay demonstrates substantial variation around integer values. Without defined gaps between each of the clusters, assignment of whole copy number is performed by rounding the data to the nearest integer. B. The MLPA assay demonstrates a distribution of 0–6 copies with peaks centered on integer values. Copy number assignment can be performed by the groups of samples to a particular cluster.

Figure 4

Comparison of the raw ratio data between MLPA, PRT and real-time QPCR using concordance plots. A. MLPA and PRT based measurements clearly demonstrates grouping around integer values and excellent correlation, as indicated by the R² value of 0.91. Five samples represent a difference between PRT and MLPA of a single copy, however clustering both PRT and MLPA measurements allows copy number to be unambiguously assigned. B. A comparison between MLPA and real-time QPCR demonstrates a greater degree of variation between the two measurements. An almost vertical spread of measurements on the concordance plot demonstrates a poor degree of clustering in the 3–6 copy range between the two assays. The correlation between real-time QPCR and MLPA is high (R² 0.87), suggesting a low level of precision with respect to QPCR.