Combining callers improves the detection of copy number variants from whole-genome sequencing

Marie Coutelier,Martin A Mensah,Peter Krawitz,Ricarda Flöttman,Marten Jäger,Dieter Beule,Denise Horn,Malte Spielmann,Manuel Holtgrewe,Stefan Mundlos

doi:10.1038/s41431-021-00983-x

Abstract

Copy Number Variants (CNVs) are deletions, duplications or insertions larger than 50 base pairs. They account for a large percentage of the normal genome variation and play major roles in human pathology. While array-based approaches have long been used to detect them in clinical practice, whole-genome sequencing (WGS) bears the promise to allow concomitant exploration of CNVs and smaller variants. However, accurately calling CNVs from WGS remains a difficult computational task, for which a consensus is still lacking. In this paper, we explore practical calling options to reach the best compromise between sensitivity and sensibility. We show that callers based on different signal (paired-end reads, split reads, coverage depth) yield complementary results. We suggest approaches combining four selected callers (Manta, Delly, ERDS, CNVnator) and a regenotyping tool (SV2), and show that this is applicable in everyday practice in terms of computation time and further interpretation. We demonstrate the superiority of these approaches over array-based Comparative Genomic Hybridization (aCGH), specifically regarding the lack of resolution in breakpoint definition and the detection of potentially relevant CNVs. Finally, we confirm our results on the NA12878 benchmark genome, as well as one clinically validated sample. In conclusion, we suggest that WGS constitutes a timely and economically valid alternative to the combination of aCGH and whole-exome sequencing.

Highlights

Structural variations (SVs) are DNA variations larger than 50 base pairs [1,2,3] and include copy number variants (CNVs), and copy number neutral variants
Whole-genome sequencing Whole-genome sequencing was performed for the probands and their parents to allow compared visual examination, but data from the index case only was used for CNV calling
The landscape of CNV calls detected by each caller is extremely variable Four callers were selected based on their ability to detect aCGH calls (Supplementary Fig. S4)

Summary

Introduction

Structural variations (SVs) are DNA variations larger than 50 base pairs (bp) [1,2,3] and include copy number variants (CNVs) (deletions, duplications and insertions), and copy number neutral variants (inversions and translocations). SVs are considered responsible for 50–95% of human samples sequence difference to the reference genome [3, 4] They are prominent in human diseases, with 15% of patients with intellectual disability or schizophrenia harboring clinically relevant CNVs [5, 6]. Molecular karyotyping relies on the simultaneous hybridization of two differentially labeled DNA samples (test and control) to an array with oligonucleotide probes and encompasses both highresolution microarray-based Comparative Genomic Hybridization (aCGH) and single nucleotide polymorphism (SNP) arrays. They reach a theoretical resolution of 1–3 kilobases (kb) with commercially available 1M arrays [12]. Whole-exome sequencing (WES) allows genome-wide identification of disease-causing coding single nucleotide variants (SNVs) and small insertiondeletions, but has limited abilities to detect larger SVs [13]

Methods

Results

Conclusion