CNVs, which occur through somatic copy number variation (CNV), play an essential role in cancer development and can be detected using various next generation sequencing technologies.
No matter the sequencing center, caller, or platform used, concordance of CNV calls between samples was consistently high due to an accurate genome ploidy estimate.
High Sensitivity
CNV detection is an integral component of genome sequencing applications such as whole genome sequencing (WGS), whole exome sequencing (WES), and targeted gene panels. CNV discovery can be an arduous task as it involves interpreting large volumes of reads with variable lengths and orientations across genomic regions of variable coverage; various NGS-based methods have been created to tackle this difficulty, such as CNVnator or read pair methods such as AscatNGS.
CNVnator stands out among RD-based methods by showing high sensitivity and the lowest FDR for detection of CNVs in both WGS and WES samples, along with more accurate breakpoint definition up to the nearest nucleotide. Furthermore, it can detect all types of CNVs, including inversions and translocations that can be difficult to identify using array CGH.
NGS-based CNV calling does have some drawbacks, however. Mapping ambiguity introduced by short read sequencing makes it challenging to accurately detect CNVs within repetitive and complex genomic regions; additionally, its sensitivity may be diminished due to variation in sequencing coverage across the genome and differing DNA input amounts.
The CNVnator algorithm employs multiple algorithms to address these challenges, including an accurate chromosomal imbalance filter, read-pair distance estimator and iterative local likelihood model to create genotype probabilistic models. After genotype identification has occurred, false positives are removed using heuristic filtering while statistical methods determine confidence levels for every CNV call, giving high quality results with low FDR rates and incorrect breakpoint assignments.
High Accuracy
Somatic copy number variants (CNVs) are an increasingly prevalent cause of cancer. These genetic mutations alter expression levels or lead to rearrangements that lead to cell dysfunction and abnormalities, with NGS technologies offering powerful ways of discovering CNVs at unprecedented speeds. With all the advancements of next generation sequencing (NGS), many CNV calling tools have been designed specifically for NGS data, with large and diverse datasets essential to assess their performance effectively.
At our facility, we utilized a curated panel of normal samples to assess the accuracy of various CNV callers, such as SNAP, gCNVkit, CODEX2, ascat and GATK gCNV. This normalization strategy relies on the concept that healthy samples produced under similar technical protocols exhibit similar patterns of random technical noise in their raw sequencing read depths; such patterns are commonly known as genomic waves.
We analyzed the effect of FFPE samples, DNA input amount, and tumor purity on CNV calling results. The FFPE process significantly reduced CNV calls in gain regions while DNA input amount and tumor purity had significant effects on caller performance; we did not find evidence that sequencing centers impacted clustering of CNVs.
Furthermore, we compared different callers’ abilities to detect large CNVs (gain > 10 kb) by comparing peak heights with thresholds used by Nexus Copy Number software 10.0 (BioDiscovery). All CNV calls were further visually inspected and confirmed using MLPA for samples GB01-GB38 and NA12878.
We found that gCNVkit, SNAP and ascat were among the most accurate CNV callers when it came to peak height accuracy in both GB01-GB38 and NA12878 samples, with ascat being superior in terms of distribution for peak height distribution in GB01-GB38 but inferior for NA12878 samples due to improved sensitivity for both gain and loss regions. Furthermore, ascat was more reliable at detecting large splice variants than either SNAP or asCNVkit and identified more large variants for NA12878 than both asCAT or SNAP while asCNVkit identified more large splice variants for NA12878 sample both with respect to gain and loss regions than either asCAT or SNAP in both gain and loss regions than either asCAT or SNAP had ever identified large variants detected by asCNVkit than either ascat or SNAP in both gain and loss regions than its counterparts which may help understand somatic CNVs as these would provide clues about potential somatic CNV mechanisms of somatic CNVs than previously identified by ascat or SNAP gCNVkit may help understand somatic CNVs by detecting more reliable detection large variants than its counterpart ascat and SNAP; Additionally it identified more large splice variants than either ascat nor SNAP in both gain/loss regions than their counterparts ascat thus potentially helping identify possible somatic CNV kit potentially identified more large splice variant detection thus becoming useful identifying somatic CNVs via its detection mechanisms than via other means of somatic CNVs detected by other callers than those reported elsewhere than before through more traditional methods than as cat while asCAT or SNAP or their presence to use than their predecessor s during somatic CNVkit identified more large CNV variants detected more large CNV variants identified more large ones than ascat/SNAP than asCAT/SNAP somatic CNV than asCAT/SNAP for NA12878 than before than as cat and SNAP detected more large variants detected s than them somatic somatically CNVs detected via these somatically, thus providing potential mechanisms that CNVs identified g CNVs detected somatic CNVs due to which may provide clues identified than they might somatic CNVs which allows when trying for somatic CNVs somatic CNVs than their predecessor SNAP/Ckit SNAP could assist s than more frequent somatic CNVs via NA12878 detected CNVs identified more large somatic CNVs!.kit identified than asCAT+NA 12878 somatic CNVs than them due. gkit may aid gkit may detecting more large detecting large SNAP would give more identified thus making. gkit than cat/SNAP thus could provide useful in somatic CNVs. gkit which may give their mechanism.
Low False-Discovery Rate
CNVs are an important source of genomic variation and influence phenotypic expression, thus contributing to disease. Their detection and characterisation is integral to genetic counselling for rare diseases as conventional techniques such as gene panel testing or microarray-based comparative genomic hybridization (array CGH) may miss some CNVs. Whole genome sequencing (WGS) offers more precise CNV discovery but requires careful evaluation in order to ensure its quality and accuracy.
This study demonstrates the value of WGS for the identification and characterisation of CNVs in patients referred to a national clinical genetics service whose diagnoses remain uncertain after standard of care investigations such as gene panel testing and microarray CGH. WGS assisted in three cases by facilitating CNV detection and confirming zygosity; ultimately leading to diagnoses for PRKN-related Parkinson disease, TAOK1-related neurodevelopmental disorder and Usmani-Riazuddin syndrome as genetic diagnoses.
WGS allows for the examination of more CNVs than other methods, including complex variants like inverted duplications and displaced duplications. Furthermore, WGS is more sensitive to repetitive sequences and can identify breakpoints at nucleotide resolution; furthermore it can provide information regarding parental origin and phasing.
We evaluated several popular CNV calling tools on WGS and WES data from FFPE tumor samples and RNA-seq tumor samples using Jaccard index clustering comparisons, and found considerable variations among them in terms of performance. AscatNgs, CNVkit, and DRAGEN proved most reliable with high concordance between WGS comparisons as well as minimal impact due to sequencing center (Additional file 1: Figure S1D-E).
WGS datasets allowed researchers to examine genome ploidy variation, which can negatively impact CNV interpretation when it deviates from expected diploid levels. This was especially evident for somatic WES samples which displayed a ploidy level of 2.85; manual re-centering of CNVkit was necessary in order to account for it; by contrast, somatic WGS dataset showed an accurate predicted ploidy level of 2.4 and was correctly predicted by CNVkit – showing how crucial it is when using short read sequencing technologies!
Easy to Operate
Next-generation sequencing (NGS) provides an accurate view of genome, which allows researchers to identify causative variants linked to complex diseases. When combined with array CGH data, NGS allows for greater insight into genetic underpinnings of disease.
Researchers who detect CNVs with higher resolution are able to investigate their role in disease, as well as detect them within samples from populations with an established incidence. This enables researchers to better identify population at risk as well as evaluate potential treatments or interventions to prevent or treat disorders more effectively.
WGS outshines array-based methods in its ability to detect CNVs in diploid samples, resolving more CNVs in complex genomic regions like gene clusters. Furthermore, WGS allows researchers to confirm or refine phase (in recessive disease cases) and parental origin more quickly as additional information from targeted regions is captured; something which would otherwise be challenging using SNP arrays or WES technology.
WGS requires only 200 ng of genomic DNA (gDNA), and can be performed using Illumina’s HiSeq 2500 or NextSeq 500 platforms. Fragments of genomic DNA are fragmented using Covaris S2 and then ligated to adaptors using KAPA HTP Library Preparation Kit before being sequenced at 50bp average read depth per lane; finally the sequence files generated are then analyzed using Illumina’s Variant Caller software.
Accurate CNV detection relies on accurately pinpointing its breakpoint position and orientation. To achieve this goal, software uses a heuristic algorithm which scans target regions for reads that align to one edge of the CNV before comparing these with reads that align to opposite edges; this enables it to precisely identify its position and orientation.
Additionally, this software also automatically detects CNVs that are located near centromeres or near genome centers and eliminates manual annotation. Finally, it analyzes data and provides a summary page displaying detected CNVs with their location within the genome; also included is an information about how many were found as well as plots showing distribution by size range.
