CONSERTING: integrating copy-number analysis with structural-variation detection

Abstract

We developed Copy Number Segmentation by Regression Tree in Next Generation Sequencing (CONSERTING), an algorithm for detecting somatic copy-number alteration (CNA) using whole-genome sequencing (WGS) data. CONSERTING performs iterative analysis of segmentation on the basis of changes in read depth and the detection of localized structural variations, with high accuracy and sensitivity. Analysis of 43 cancer genomes from both pediatric and adult patients revealed novel oncogenic CNAs, complex rearrangements and subclonal CNAs missed by alternative approaches.

Figure 1

Strategy for CNA detection used by CONSERTING. CNAs are identified through iterative analysis of (i) local segmentation by read depth (RD) within boundaries identified by structural variation (SV) breakpoints followed by (ii) segment merging and local SV analysis. Yellow vertical bars mark SV breakpoints. Dotted boxes indicate the candidate breakpoint regions for local SV analysis, which display RD changes but are not reported in global SV analysis.

Figure 2

Comparison of WGS CNAs detected by CONSERTING and four other methods. (a) A Circos plot which displays CNAs found by all six methods in one of the 12 ETP-ALL samples, SJTALL007. (b) Box plots showing F₁ scores of WGS CNAs (compared against CNAs curated from SNP arrays) and the number of CNA segments uncorroborated by SNP arrays in the 12 ETP-ALL samples. The box represents the interquartile range (IQR) while whiskers extend to the most extreme data point which is no more than 1.5-fold of IQR away from the box. (c) Box plot of F₁ score of WGS CNAs and SNP-array CNAs and number of CNA segments uncorroborated by SNP array analysis in the 22 TCGA-GBM samples. (d) A Circos plot which displays CNAs found by CONSERTING and BIC-seq in the diluted COLO-829 (scale adjusted for dilution effects), published COLO-829 CNA from un-diluted COLO-829 and SKY map data. The 1-copy gain of chromosome X was found only in the diluted sample by both CONSERTING and BIC-seq, which is consistent with the SKY data as there are 2 chromosome X in this cell line derived from a male patient.

Figure 3

A complex re-arrangement in a pediatric low grade glioma sample identified by CONSERTING. (a) Circos plot of 8 chromosomes (1, 3, 4, 10, 11, 12, 16 and 22) with inter-chromosomal SVs (purple lines) and intrachromosomal SVs (green lines) connecting the amplification CNAs (red dots). (b) SV graph constructed from CNAs and SVs identified on chromosomes 3, 11 and 12. The black lines indicate SVs detected only by CONSERTING. The red, blue and green dots mark the three BAC clones selected for FISH assay. The SV represented by red-blue fusion represents an in-frame FYCO1-RAF1 fusion. (c) Red, green, blue fusion signal found in 36% of the nuclei. d) Various 2-color fusion signals including red-blue fusion and red-green fusion (scale bar: 10 μm).