CNV

The copy number variant caller in the TSO 500 ctDNA analysis software performs amplification, reference, and deletion calling for 59 genes. The list of genes can be accessed on the TSO 500 ctDNA product page.

CNV calling algorithm of the TSO 500 ctDNA analysis software

Target Counts

The first step in the CNV calling algorithm identifies target counts. It includes extracting signals such as read count and improper pairs and putting them into target intervals.

GC Bias Correction

GC bias refer to systematic differences in read coverage that are related to the GC content of the DNA (percentage of guanine (G) and cytosine (C) bases). Library preparation, capture kits, sequencing system differences, and mapping can contribute to GC bias (Note: many of these variables are also captured in the Panel of Normals covered in Normalization below). Regions with very high GC content might be underrepresented because they are harder to amplify or sequence. Regions with moderate GC content might be overrepresented.

GC bias can impact CNV calling. The DRAGEN GC bias correction module is designed to correct GC bias and generate GC-bias-corrected target counts.

Normalization

The goal of normalization is to correct for technical biases and artifacts introduced during sample preparation, sequencing, and data processing. These biases can otherwise lead to reductions in sensitivity and specificity in CNV calling.

In the CNV caller in the TSO 500 ctDNA analysis software, the normalization step uses a panel of normals (PON) to determine the baseline level from which to call CNV events. Illumina carefully assembles the PON to include a cohort of healthy cell-free DNA samples and contrived samples and updates the PON as needed to improve performance. The PON samples are processed using either TSO 500 ctDNA v1 or v2 library prep kits to accompany different library prep kits within the portfolio.

The normalized read counts produced at this step are then used for segmentation and fold-change calculation.

Targeted Segmentation

In context of CNV calling, segments are regions of the genome with similar read-depth or copy number profiles. The TSO 500 ctDNA analysis software does not use a segmentation algorithm to assess CNV events, but rather they are derived from a pre-defined CNV resource BED file included with the software. The boundaries within this file determine the intervals used for copy number variant analysis.

Fold-Change (FC) Calculation

Fold change measures how much the read coverage in a given genomic region, for example, containing a target gene, differs from a baseline.

The CNV caller in the TSO 500 ctDNA analysis software calculates Fold Change using a formula:

$$Fold Change = Observed Coverage/Expected Coverage$$

• Observed Coverage: The actual read depth in a region.

• Expected Coverage: The average read depth in regions assumed to be diploid (2 copies) as represented in the panel of normals (PON).

Calling / Genotyping

The CNV caller in the TSO 500 ctDNA analysis software uses gene-specific thresholds established by Illumina based on the PON to call amplifications and deletions. The thresholds indicate what fold-change values each gene needs to reach to be called an amplification or a deletion.

In the VCF notation, <DUP> indicates that the detected fold change is greater than a predefined amplification cutoff for the that gene. <DEL> indicates that the fold change is less than a predefined deletion cutoff for that gene. The cutoffs vary from gene to gene and stored in cnv/dragen_tso500_manifest_59genes_predefined_cutoff_cfDNA.bed.

Output

CNV calls are output in several files:

• Combined Variant Output File, {Sample_ID}_CombinedVariantOutput.tsv

• Copy Number Variants VCF , {Sample_ID}.cnv.vcf

• Copy Number Variants Metrics CSV, {Sample_ID}.cnv_metrics.csv

Combined Variant Output File

File name: {Sample_ID}_CombinedVariantOutput.tsv

The Combined Variant Output File displays CNV calling results in the section [Copy Number Variants]. Each gene is accompanied by a fold change value.

To be included in the Combined Variant Output File, copy number variants must meet the following conditions:

• FILTER field in the Copy Number Variants VCF file marked as PASS

• ALT field is <DUP or <DEL>

• Gene is part of the copy number variant gene list

Copy Number Variants VCF

File name: {Sample_ID}.cnv.vcf

The file includes CNV calling results in the VCF format. <DUP> indicates amplifications, <DEL> indicates deletions.

Each CNV call is accompanied by a Q-score in the QUAL column. Q-score is a Phred transformation of the p-value represented by the following equation, where p-value is derived from the t-test between the fold change (FC) of the gene against the rest of the genome:

Higher Q-scores indicate higher confidence in the CNV call.

The Copy Number Variants VCF includes additional details on CNV genomic position, breakpoints, bins and segments.

Copy Number Variants Metrics CSV

File name: {Sample_ID}.cnv_metrics.csv

The file includes copy number variant metrics reported on a per sample level.

Name	Description
Sex Genotyper	The predicted sex of the sample
Bases in reference genome	The number of bases in the reference genome
Average alignment coverage over genome	The average alignment coverage across the reference genome
Number of alignment records	The number of alignment records in the sample
Number of filtered records (total)	The number of total filtered records
Number of filtered records (duplicates)	The number of duplicated filtered records
Number of filtered records (MAPQ)	The number of MAPQ filtered records
Number of filtered records (unmapped)	The number of unmapped filtered records
Coverage MAD
Gene Scaled MAD
Median Bin Count
Number of target intervals	The number of target intervals in the sample
Number of normal samples
Number of segments	The number of segments in the sample. Applicable only to CNV SLM mode
Number of amplifications	The number of amplifications in the sample. Applicable only to CNV SLM mode.
Number of deletions	The number of deletions in the sample. Applicable only to CNV SLM mode.
Number of passing amplifications	The number of passing amplifications in the sample. Applicable only to CNV SLM mode.
Number of passing deletions	The number of passing deletions in the sample. Applicable only to CNV SLM mode.