HTS Bioinf - ELLA CNV module: user interface (UI)

Scope

This document describes the user interface (UI) of the CNV module, including navigation, variant display, classification workflow, and annotation features for effective CNV analysis and interpretation.

Description of the CNV module UI

Indication of CNV inclusion in an analysis

To confirm if CNVs are included in an ELLA analysis, check the sample in the overview. A single HTS tag indicates only small variants are reported, while two tags confirm CNV inclusion. Alternatively, within an open analysis, navigate to the info field: a single record means no CNVs, while double records indicate their presence.

Figure 1: HTS tags indicating small variants only (single) or small variants with CNVs (double).

Figure 2: Info section samples to check for CNV analysis. The info field within an open analysis displays either a single record (indicating no CNVs) or two records (indicating the inclusion of CNVs) for the sample.

CNV results view

CNV results are available by clicking the “CNV” button in the top banner, as shown in Figure 3:

Figure 3: Button to switch between the SNV and CNV modules.

The sidebar

The sidebar (highlighted in Figure 4) lists all called CNVs that have not been hard-filtered (e.g. removed) based on the parameters described in ELLA CNV module: CNV calling and filtering or soft-filtered (hidden in the filter tab in ELLA), more details below.

Figure 4: CNV module with the sidebar highlighted.

The sidebar is divided into sections, similar to the SNV module. Unclassified variants lists variants with no assigned classification in ELLA's database, while Classified variants lists those that have been classified either during the current review/interpretation session or previously (provided the variant is 100% identical to a previously evaluated variant). In addition, ELLA filters out any variants (100% identitical) set at AMG to Class 1 or 2 where the classification is still valid. These will be "hidden" behind the Filter button.

Note

Previously classified variants that overlap but are not identical to the current variant under review will not be defined as classified but will instead appear under Unclassified variants. To identify overlapping CNVs in the same region, you must use the Visual module with the "Classified" track or one of the aCGH tracks.

Columns in the sidebar

• SVTYPE: Specifies the type of predicted CNV. Possible values:
  • DEL: Deletions.
  • DUP: Duplications identified by the CNV caller Canvas. The region indicated in DUP specifies the genomic region duplicated but does not specify the destination within the genome. Raw data must be manually reviewed in IGV to assess this.
  • DUP:TANDEM: Duplications identified by the SV caller Manta.
• CHR: The chromosome to which the CNV is mapped.
• POS: Chromosomal position (start-stop) of the CNV based on the GRCh37/hg19 reference genome. The position is defined by the CNV call, but manual inspection of raw data in IGV is necessary to determine whether this reflects the actual CNV breakpoint. By default, CNVs in the sidebar are sorted in ascending order by CHR and then by POS.
• GENE: Genes in the relevant gene panel that overlap with the CNV. Note that only genes in the gene panel are listed, so it is essential to consider that the CNV may include more genes than those listed here. Larger CNVs may also encompass multiple genes in the gene panel; hover over the gene name to view these.
• BAND: The chromosomal band(s) overlapped by the variant.
• LENGTH: The length of the predicted CNV, reported in bp or kb depending on its size. Note that variants >1000 bp are converted to kb in the sidebar list, but mouse hover or looking in the top banner reveals the full/exact length. The column can be sorted by actual length by clicking on the header.
• HGVSg: A short version of the variant name in HGVSg format (GRCh37). The full HGVSg annotation is displayed in the top panel.
• CSQ: The predicted consequence in transcript regions, provided by VEP (as for SNVs), but limited to CNVs < 10 Mb (due to performance limit). The effect on protein-coding sequences must be manually verified using IGV.
• CLASS: Displays the variant's classification in ELLA, including the class from any previous review/interpretation session (provided the variant is 100% identical to a previously evaluated variant), similar to SNVs.

Top banner

The top banner displays genes included in the current gene panel. Note that the CNV may include other OMIM-morbid genes (and other genes) that are not shown here.

Figure 5: CNV module with the top banner highlighted.

Annotations in the top banner:

• 1: Inheritance: As defined for the gene in the current gene panel (as for SNVs).
• 2: Gene: The relevant gene with a pop-up containing detailed information (as for SNVs). Transcript annotations for the gene are available under GENE INFO and in the "gene panel" track in Visual.
• 3: HGVSg and genotype: The variant name in HGVSg format (GRCh37) and genotype. For duplications (DUPs), the copy number is displayed instead of the genotype. For all other CNVs, the allele is represented as the number of nucleotides in parentheses.
• 4: HGVSp: The protein change in HGVS nomenclature, provided by VEP (as for SNVs). This field is most valuable for intragenic variants, as predictions are not designed for variants spanning multiple genes and are often empty for CNVs. The annotation only accounts for the region identified as affected by the duplication or deletion (not, for example, the location of the duplicated material). Manual inspection of raw data is necessary to assess the CNV's effect on the affected protein.
• 5: CSQ: Same as in the sidebar.
• 6: Chromosome and position: For the CNV in question, the same as in the sidebar.

Classification (FULL mode)

By default, variants are displayed in FULL mode, which is structured similarly to the SNV module. Parameters in the full window are described below:

Section: Quality

• Filter: Variants labeled PASS are not filtered out by upstream frequency filters in ELLA. Variants marked HiFreq in the normal databases may still appear. This includes homozygous variants on X (HomDELX) and AnnotSV class 4 and 5 variants (HighACMGClass). These are retained after frequency filtering. The minimum solution does not provide standardized quality metrics for CNV calling confidence. CNV quality is assessed based on changes in read depth, the number of improperly paired reads, and the number of split reads. These metrics are found in the CNV VCF track in the Visual module and can support visual observations. Remaining values under "Analysis-specific quality" should be used cautiously:
• Quality: A score summarizing the impact of parameters such as read depth, improper read pairs, and split reads. Higher values indicate better quality but are conditional on the caller used, the variant's size, type, and genomic location.
• GQ: Indicates the confidence in genotype accuracy, with similar caveats as Quality.
• Depth and Ratio: Currently, these fields have currently no values for CNVs in ELLA.

Section: Classification

When a CNV is classified as class 3, 4, or 5, it will be listed in the 'Report' tab with both HGVS and ISCN annotation.

CNVs classified as class 1 or 2 will be filtered out in the next patient, provided 100% identity and still valid.

Note

The suggested ACMG criteria have not been sufficiently tested for CNVs and should therefore be ignored.

Section: Region

Lists the 10 nearest classified variants (in each direction) from the ELLA database, dynamically updated when a variant is selected in the sidebar.

Sections: Frequency and External

CNV assessments should primarily occur in the Visual module, as comparison data in these sections is limited.

Visual Mode

Tracks can be toggled on or off either via PRESETS or individually under TRACK SELECTION. Tracks can be removed individually even if a PRESET has been selected initially.

Note

Please note that when the BAM file is visualized in visual-IGV with "color by: pair orientation and insert size," insert size will override pair orientation. This differs from desktop-IGV, where pair orientation overrides insert size, which is the correct display. This issue is due to a bug in the version of IGV used in visual-IGV. The bug has been reported and is awaiting a fix from IGV.

Track presets

Track presets select all tracks grouped under the corresponding header in Track selection. The tracks in each preset are described in the tables below.

Table 1: Core track presets overview

Track	Description	Default	SNV	CNV Default	Raw	Gene/Region	Control
VARIANTS	SNVs and CNVs for the patient (filtered).	+	+	+
ROI	Vertical green marker field for all variants in VARIANTS track.	+
CNV VCF	Raw CNV data with all called CNVs, including annotations.	+		+	+
REFSEQCURATED	Gene names and curated transcripts (NM_/NR_). The track is collapsed by default, but you can expand it using the gear icon to view individual transcripts, or click on the track to see all overlapping transcripts in a pop-up.	+	+	+		+
GENE PANEL	Includes only genes in the gene panel, with the default transcript used for the sample analyzed.	+	+	+		+
CLASSIFICATIONS	All classifications for all patients/variants analyzed in ELLA, dynamically updated.	+	+	+
GENE-BASED	Combined track for genes from various data sources*, green for OMIM morbid genes, otherwise black.	+	+	+		+
COVERAGE: CONTROL NA12878	Sequence coverage (bigWig) for control sample NA12878.			+			+
BAM: PROBAND	Sequence alignment for the proband (and parents, if available, in separate tracks). Only visible up to zoom-level of 5kb.	+	+	+	+
COVERAGE	Sequence coverage (bigWig) for proband (and parents, if available, in separate tracks), showing low resolution up to chromosome-level.	+		+	+
BAM (HC)	Displays the "haplotype caller" BAM file for proband, showing sequence coverage only in regions where variant calling has performed a local re-alignment				+
BAM: CONTROL NA12878	BAM file from a control sample NA12878. The WGS run can be viewed by hovering over the "button" in the track selection Control or the track label in the IGV browser.		+	+			+
GIAB HIGHCONF	Regions with 'high confidence calls' in 'Genome in a Bottle' (GIAB) NA12878.						+
SNV VCF	Raw data with all called SNVs, including those filtered out.
SEGDUP	Segmental duplications from genomicSuperDups, with a color scale showing homology degree.					+
CLINGEN CURATED	Recently curated regions from ClinGen with HI scores, ISCA, cytobands, and TS scores. These correspond to the 500+ regions described in ClinGen, plus LCR regions, often referred to as breakpoints (BP)					+
UNIPROT DOMAINS	Protein domains from UniProt.					+

---

*GENE-BASED track details. Combined track from the following data sources:

• OMIM:
  • pheMIM: OMIM-ID for related phenotypes
  • phenotype: Phenotype from OMIM
  • inheritance: Inheritance pattern from OMIM
  • Morbid gene: yes/no
  • HI and TS: From ClinGen. Full evidence report available in ClinGen. The score shown is:
    • NA = not evaluated yet
    • 30 = recessive condition
    • 40 = unlikely to be dosage-sensitive
    • 0 = no evidence
    • 1 = limited evidence
    • 2 = emerging evidence
    • 3 = sufficient evidence
• GenCC: Link to GenCC curated gene-phenotype relationship page.
• HGNC: Gene identification in the HGNC database
• pLI: Related to pLoF from gnomAD; indicates the gene's likelihood of intolerance to Loss-of-function (LoF) variants based on its underrepresentation of LoF in gnomAD. Genes with high pLI (close to 1) are LoF intolerant.
• LOEUF: Related to pLoF from gnomAD; stands for Loss-of-function Observed/Expected Upper bound Fraction, and indicates the observed underrepresentation (or enrichment) of LoF variants in gnomAD.
• DDG2P: Shows genes with DDD annotation in DECIPHER (yes/no).
• perc HI: % HI from DECIPHER; low percentages indicate a high probability that the gene is haploinsufficient.
• ouslist: Internal AMG lists with genes that have special exceptions for imprinting/mosaicism in SNV interpretation.

Note

It is recommended to use ClinGen or Decipher directly for regions with multiple genes or when more detailed, consolidated information is needed.

Table 2: Population and clinical databases presets for structural variants

Track	Description	CNV Pathogenic	CNV VUS	CNV Benign
GNOMAD SV	All structural variants from gnomAD v2.1.1.			+
GNOMAD SV CONTROLS	Subset of gnomAD SV, only variants from individuals not used in case-control studies.			+
GNOMAD SV NON-NEURO	Subset of gnomAD SV, only variants from individuals without neurological phenotypes.			+
DGV GOLD	Curated subset of DGV, CNVs found in healthy controls across multiple publications, primarily aCGH, as well as 1000Genomes and GnomAD SV from 2019. To examine references, visit UCSC			+
INDB	AMG's in-house database showing variant frequency, type, and size. Displays Manta frequency when available, otherwise Canvas frequency. May differ from annotated frequency.			+
OUS AMG CLASS 1	Patient findings* (red: dels and blue: dups) classified as Class 1 by aCGH.			+
CLINGEN PREN CLASS 1+2	Patient data from ClinGen prenatal (nstd75). Blue: gain, red: loss.			+
CLINVAR CLIN SV CLASS 1+2	Patient data from ClinVar Clinical Structural Variants (nstd102).			+
CLINGEN PREN CLASS 3	Patient data from ClinGen prenatal (nstd75). Blue: gain, red: loss.		+
CLINVAR CLIN SV CLASS 3	Patient data from ClinVar Clinical Structural Variants (nstd102).		+
OUS AMG CLASS 4+5	Patient findings* (red: dels and blue: dups) classified as Class 4, 5, or 5R by aCGH.	+
CLINGEN PREN CLASS 4+5	Patient data from ClinGen prenatal (nstd75). Blue: gain, red: loss.	+
CLINVAR CLIN SV CLASS 4+5	Patient data from ClinVar Clinical Structural Variants (nstd102).	+
DECIPHER SYNDROMES	Known 66 microdeletion/duplication syndromes from Decipher.	+

---

Note

*OUS AMG aCGH findings: Named, for example, "Class:5:loss". Class 5R refers to findings in recessive genes that could not explain the disease on their own. The terms "Loss" and "Deletion", as well as "Gain" and "Amplification", are used interchangeably, as the data is taken directly from the analysis software where the thresholds for "Loss/Deletion" and "Gain/Amplification" are set at log2-ratio -1 and 0.58, respectively. Variants near these thresholds may have been called as both "Loss" and "Deletion" (or "Gain" and "Amplification"). The track shows only individual findings (one entry corresponds to one patient), so it cannot be used for frequency calculations.