Reference-unbiased copy number variant analysis using CGH microarrays

Abstract

Comparative genomic hybridization (CGH) microarrays have been used to determine copy number variations (CNVs) and their effects on complex diseases. Detection of absolute CNVs independent of genomic variants of an arbitrary reference sample has been a critical issue in CGH array experiments. Whole genome analysis using massively parallel sequencing with multiple ultra-high resolution CGH arrays provides an opportunity to catalog highly accurate genomic variants of the reference DNA (NA10851). Using information on variants, we developed a new method, the CGH array reference-free algorithm (CARA), which can determine reference-unbiased absolute CNVs from any CGH array platform. The algorithm enables the removal and rescue of false positive and false negative CNVs, respectively, which appear due to the effects of genomic variants of the reference sample in raw CGH array experiments. We found that the CARA remarkably enhanced the accuracy of CGH array in determining absolute CNVs. Our method thus provides a new approach to interpret CGH array data for personalized medicine.

Figure 1.

Detection of NA10851 CNVs using conjugative methods, massively parallel sequencing and ultra-high resolution CGH arrays. (a) Distribution of RD of coverage of NA10851 sequencing for CNV segments identified from CGH arrays of 73 individuals. (b) Identifying CNVs of NA10851 using RD of sequence coverage on putative regions determined by CGH arrays. (c) Examples of four categories of candidate CNVs by visual inspection. Top row, Apparent CNV; second row, Indistinct region; third row, Indeterminable region due to extremely unstable RD; fourth and bottom, Nested CNV, the CNV candidate in the bottom row is removed since it is included in the CNV illustrated in the fourth row.

Figure 1.

Detection of NA10851 CNVs using conjugative methods, massively parallel sequencing and ultra-high resolution CGH arrays. (a) Distribution of RD of coverage of NA10851 sequencing for CNV segments identified from CGH arrays of 73 individuals. (b) Identifying CNVs of NA10851 using RD of sequence coverage on putative regions determined by CGH arrays. (c) Examples of four categories of candidate CNVs by visual inspection. Top row, Apparent CNV; second row, Indistinct region; third row, Indeterminable region due to extremely unstable RD; fourth and bottom, Nested CNV, the CNV candidate in the bottom row is removed since it is included in the CNV illustrated in the fourth row.

Figure 2.

Personal CNVs of NA10851. (a) Personal CNV distribution throughout the entire genome. (b) Numbers and lengths of CN losses and CN gains of NA10851. (c) Size distribution of 1309 NA10851 CNV regions. (d) Repetitive context of CN gains and losses.

Figure 3.

Flowchart showing the CARA for detection of reference-independent CNVs using CGH arrays.

Figure 4.

Evaluation of the utility of CARA. (a) Comparisons between CNV sets of AK1 before and after application of CARA. (b) Alterations in CNV segments upon application of CARA. (c) Concordance between CGH arrays and RD of sequence coverage after application of CARA.