The variant page top bar displays, from left to right:

The Related Cases tab highlights tagged variants found in previously analyzed cases, both within your organization and among organizations in your network(s).

Key features:

1. —enable viewing of network-wide case data alongside internal case data

2. —visualizes the distribution of classification outcomes for a variant across previously analyzed cases

3. —allows to dynamically filter CNVs by overlap percentage

The Clinical Significance tab provides essential variant-level and gene-level details to help you evaluate a variant’s potential pathogenicity. It combines AI-driven insights with curated public data for informed interpretation.

Information is organized into the following cards:

• : Displays key details such as HGVS nomenclature, predicted effect, variant type, zygosity, and links to external databases.

• : Shows individual tool results and algorithmic predictions for missense, conservation, and splicing effects.

The Variant info card provides key details about the variant.

Details included

• Variant type

The Resources field provides links to external databases with pre-filled search queries, helping users efficiently gather supporting evidence for variant interpretation.

Available resources

• UCSC Genome Browser

• GeneCards (retired in v100.39.0)

• PubMed

• WikiGenes (retired in v100.39.0)

• Google Scholar (v100.39.0+)

• LitVar2 (v100.39.0+)

• Genomenon

How search queries are built

v100.39.0+ and above

Search queries are built using detailed variant information whenever possible.

The smart query builder:

• Supports variant aliases

• Takes variant type into account for more precise and context-aware results

The query logic varies by variant type:

• SNVs/indels (including mtDNA): Includes gene symbol, dbSNP identifier, HGVS coding change, and HGVS protein change across affected transcripts

• CNVs: Includes gene symbol, ISCN notation, cytoband, genomic location, and standard terms like «CNV», «deletion» and similar

• SV insertions: Includes gene symbol, cytoband, genomic location, and standard SV descriptors, and standard terms like «SV», «rearrangement» and similar

Versions prior to v100.39.0

Search queries are based on gene information only.

The Variant tagging widget now supports assigning multiple tags to a single variant for every user, making tagging more flexible and expressive. This allows you to mark variants for different purposes without losing clarity. For example, you can tag a variant as both “Candidate” (for primary findings) and “Incidental” (for secondary findings reporting).

This feature is controlled at the organization level. Once enabled in your organization’s settings, it applies to all new and in-progress cases. If disabled, tagging reverts to the original model, where only one tag could be applied per variant.

How multiple tags work

Tagging a variant

• In the variant page header, the tag dropdown becomes a multi-select menu.

• Click to choose one or more tags. Each tag you select is added to the variant.

• Your avatar icon appears next to the tag, showing that you assigned it.

Viewing tags

• All tags assigned to the variant are displayed in the page header.

• In the dropdown, you can see which users assigned each tag by their avatar icons.

• At the bottom of the dropdown, you’ll also see a list of users who have viewed the variant.

Removing tags

• You can remove your own tag by clicking the red X next to your avatar.

• Tags assigned by others remain intact unless you have a supervisory role, which allows removing tags assigned by other users.

Tag visibility in cases

• In the Related Cases table, all tags assigned to a variant are displayed.

• If both user tags and AI tags exist, only user tags are shown in the variant header for clarity.

• The case interpretation view shows only the tags used in that case, along with the number of variants per tag.

Activity tracking

Every tag assignment or removal is logged in the case activity history, including:

• User name

• Tag name

• Variant details

• Timestamp

Example entry: “[User] removed tag [Candidate] from Variant chr1:123456 A>G”.

The Transcript field displays variant description according to the Human Genome Variation Society (HGVS) standard for coding DNA and protein changes.

Some transcripts are marked with one or both indicators:

• Checkmark : Canonical transcript

• Curate icon: Transcript selected in your database.

Emedgene automatically selects a reference transcript according to predefined . The reference transcript determines:

• Main affected gene

• Main variant effect

• Coding DNA change and resulting protein change

Changing the reference transcript

If the automatically selected transcript does not match your reporting preference, you can:

• Select a different transcript* from the Transcript dropdown menu

• Add a custom transcript as part of a custom HGVS variant description (see below). You may need to add a custom HGVS variant description in the following cases:

  • The required transcript is not available in the suggested options

How to add a custom HGVS variant description

Once added, a custom description becomes available for .

Variant tags are shown in the Variant tagging widget of the Variant page Top bar.

When a variant is tagged by a user, the user's tag is displayed in the tag field. If the variant is only tagged by AI Shortlist and not by a user, the tag AI Shortlist is displayed. If a variant has not been tagged, the tag is shown as N/A.

A variant can be tagged by the AI Shortlist or a user as:

• Most Likely Candidate - most promising for solving the case

• Candidate - worth considering

• - meets the ACMG criteria for reporting secondary findings

• Carrier - identified by the Carrier analysis pipeline

• In Report - manually selected to be reported

• Not relevant - automatically tagged variant that has been disregarded after manual review

• Any custom tag used in your organization (e.g., Submitted for Sanger confirmation)

Variant tagging widget how-to

Add or change a tag

Click on the dropdown icon in the Variant tag field and select a suitable tag. The Variant tag field showcases the most recently assigned tag.

Review the tagging history

Click on the Variant tag field and scroll to the Assigned tags section.

Remove an automatic tag

Click on the dropdown array in the Variant tag field and select Not relevant.

Clear your manual tag:

Click on the dropdown array in the Variant tag field and select Clear.

The Gene metrics card helps you assess whether a gene is tolerant or intolerant to variation. It includes:

• Intolerance metrics from ExAC and gnomAD that relate to gene constraint and variant burden

• Direct links to ExAC and gnomAD for quick access

The Connected variants tab introduced in v36.0 reveals connections between the variant under review and other variants in a case. This information can be crucial for assessing variant pathogenicity.

Supported variant types: SNVs, indels, MNVs, SVs, CNVs, mtDNA variants, and STRs.

The number of connected variants appears in parentheses after the tab title (for example, "Connected Variants (3)").

The Population Statistics tab addresses detailed population allele data across various ethnicities in public and internal databases. Field labels reflect their source databases—such as gnomAD, ExAC, and internal repositories—making it easier for users to trace allele frequency data back to its origin.

• Public databases (SNVs): 1000 Genomes, ESP 6500, ExAC, and gnomAD

• Public databases (CNVs): 1000 Genomes, gnomAD SV, Decipher, DGV

The Network data view controls enable users to display network-wide case data alongside internal case data, providing a broader context for variant interpretation.

These controls are located on top of the Related cases tab.

Display modes

There are two display modes for the Network data view: Simple mode and Advanced mode. Each mode provides a different level of control depending on your needs.

The Variant page > tab offers a Dynamic CNV overlap percentage filter as well as an Overlap column for CNVs in the data table. This percentage is computed by dividing the length of a common region between CNVs by the size of the CNV under study.

The dynamic CNV overlap percentage filter, located at the top of the Related Cases tab, allows users to manually adjust the lower threshold for between the current CNV and previously reported (candidate) CNVs using a slider.

A candidate CNV is included in the list only if overlap percentage (overlap%) meets or exceeds the threshold set by the slider. In simpler terms: at least that proportion of the candidate CNV must lie within the coordinates of the current CNV.

The expandable Variant page sidebar provides access to key case information, integration with Alamut and IGV, and a detailed variant activity log—all within the Variant page.

To expand the sidebar, click the arrow icon in the top-right corner. Click it again to collapse the sidebar.

The sidebar includes three tabs:

Case info

The Case info tab displays sample names, BAM file locations (if available), affected vs. healthy sample status, and proband phenotypes.

Applications

The Applications tab lets you enable or disable connections with your IGV and Alamut desktop applications. This means there won't be any IGV and Alamut windows popping up unless you want them to!

Once you configure your connection preferences, they are saved to your user account and applied across all cases.

Activity

The Activity tab is your complete audit trail for a variant. It records variant-level user actions, including:

• Tagging variants

• Adding comments

• Drafting interpretation notes

• Editing the evidence graph

The timestamped logs support collaboration and maintain a traceable record of variant interpretation.

The Clinical significance card provides insights into how a variant has been classified previously—both within your organization and across public databases. This helps you quickly assess existing interpretations and decide whether additional evidence is needed.

Pathogenicity classification sources

• Manually classified

  Indicates if the variant was classified in any previous case in the account

• Networks classified

  Shows if the variant was classified by partner organizations in your

• Curate Displays whether the variant exists in your variant database

• ClinVar

  Lists ClinVar submissions for the variant

• ClinGen Regions (CNVs) Indicates whether a variant overlaps a dosage-sensitive region defined by ClinGen

The Pathogenicity trend bar displays the distribution of previously attributed pathogenicity classifications and enables filtering of the based on a selected pathogenicity class.

Color legend

• Green

The Visualization tab provides access to an -based variant visualization tool that allows users to review alignment and annotation data in support of variant validation and interpretation. The viewer is embedded directly in the Visualization tab and can also be accessed from the (v100.39.0+).

For a detailed description of the viewer's features, controls, and supported file formats, refer to section of the manual.

One-way CNV annotation overlap measures how much of a candidate CNV (previously reported) is covered by the current CNV.

This value is displayed in the Overlap column of the Related cases table and is used by the CNV overlap percentage filter in the Related Cases tab to refine case results based on a user-defined minimum overlap threshold.

How overlap is calculated

The Gene-related diseases card lists known disease associations for the gene or genes affected by the variant. If the variant affects multiple genes, the list shows associations for each gene.

Information displayed

Each disease entry includes:

• Disease name

  As defined in gene-disease connection source(s)

• Source links:

  • OMIM

Disease entry may additionally include:

• Monarch link (v100.39.0+) Redirects to the disease entry in the Mondo Disease Ontology supported by The Monarch Initiative.

• GenCC badge (v100.39.0+) Diseases with one or more entries in the GenCC (Gene Curation Coalition) database display a badge next to the disease name. This badge represents the highest-level gene–disease connection validity classification. If multiple GenCC entries exist for the same gene–disease connection, a “+n” indicator appears beside the badge. Hovering over this indicator reveals the additional classifications and their sources.

Disease selection

Once a variant is tagged, a default disease is automatically selected. Users have the following options:

• Leave the default disease

• Select a different disease from the available list of diseases

• Add a custom disease name

The selected disease appears in the Gene-related diseases card of the , as well as on the , and can be included in the .

Selecting a different disease

The evidence graph will update automatically to reflect the new disease association.

The change will be recorded in the Activity Log.

Adding a custom disease

Custom disease name will be stored in the case only. If you export the variant to Curate, no disease will be saved in Curate for that variant.

The Related cases table provides quick access to variant and case-specific details of previous variant interpretations, both within your organization and among organizations in your . The table shows key data at a glance, with more detailed information available when you click on a row.

Related cases table columns

Where can I review and edit the ACMG Classification of a SNV?

The ACMG SNV Classification wizard is located in the Evidence tab of the Variant page. It facilitates classification of variant pathogenicity through the automation of 26 out of 28 ACMG criteria and enabling manual review and editing of the tags presented as interactive buttons.

The ACMG SNV Classification Wizard is located in the Evidence section of the Variant Page and is designed to guide users through the interpretation of single nucleotide variants (SNVs) using the ACMG/AMP framework. It automates 26 out of 28 ACMG criteria, with PS4 requiring full manual entry and BS2 partially automated (50% manual).

The ACMG SNV Classification wizard includes a pathogenicity bar that visually represents the .

The wizard is available for tagged SNVs in disease-associated genes and displays a visual pathogenicity bar summarizing the cumulative pathogenicity score.

The wizard is available for tagged sequence variants in disease-associated genes. The results of the classification are also highlighted in the of the . Unlike the wizard, automatically assigned criteria and resulting variant class are shown in the for all variants in disease-associated genes, regardless of their status.

How do I review and edit the ACMG Classification of a SNV?

Each ACMG tag is represented by an interactive button including a checkbox for selection (1), the criterion name (2) and evidence strength indicator (3).

Pathogenic criteria are represented by red boxes, while benign criteria boxes are colored green. Each ACMG criterion has three possible states:

• Neutral (1) - represented by an empty checkbox. Criterion requires further investigation.

• Negative (2) - represented by a cross. Criterion is not applicable.

• Positive (3) - represented by a tick and dark color. Criterion is applicable.

Each ACMG tag can be manually checked, unchecked, or set to an undefined state by clicking the interactive button's checkbox element.

To examine in detail or modify the underlying evidence for the particular ACMG tag, select it by clicking on the tag name. The button becomes flood-filled (b), as opposed to it's original, non-selected, state (a).

Upon selection, a description of the criterion and its underlying evidence emerges below. Yes and No radio buttons accompany each piece of evidence. The tag can be assigned if Yes has been selected for all the underlying conditions.

You may modify evidence strength in the Strength dropdown (Stand Alone, Very Strong, Strong, Moderate, Supporting), which will impact both the pathogenicity class and score calculations.

Notes can be added to any tag, and changes are saved using the Save button.

After you've modified ACMG classification, you can either save manual changes by pressing the Save button or reset via Revert manual changes. Keep in mind that after saving your edits, Revert manual changes will become unavailable.

Warnings for ACMG tag schema versions

ACMG classifications rely on a defined ACMG tag schema, which is periodically updated to reflect new ClinGen/ACMG recommendations or platform logic improvements. To ensure classification accuracy, schema versioning is now tracked independently from the overall variant analysis engine.

When opening a case in Analyze, the system compares:

• The ACMG schema version currently applied in the case analysis

• The ACMG schema version used when the variant was curated in Curate

If the curated schema is older than the current schema:

• Curated ACMG tags will not be applied in Analyze

• The system reapplies the current schema’s criteria

• A warning appears:

You can synchronize schemas by clicking Update Curate on the variant. This updates the variant’s ACMG schema version in Curate to match Analyze, and the updated classification is saved like any other variant update.

Disease source used in ACMG evaluation

When a gene–disease association is available from multiple databases (for example: OMIM, CGD, Orphanet etc.), the variant page and gene related diseases card display all available disease sources. However, ACMG inheritance-based rules use only the OMIM disease to determine the inheritance mode.

• Multiple disease entries may appear in the UI.

• ACMG logic selects only OMIM entry when determining the inheritance mode.

• If OMIM does not include a pre-populated inheritance mode, ACMG rules that rely on inheritance mode will not be triggered—even if another source (such as CGD) contains this information.

ACMG classification of mtDNA variants

The ACMG SNV Classification wizard is available for ACMG classification of tagged mtDNA variants. To classify an mtDNA variant, please manually assign the relevant criteria; the resulting ACMG classification will be calculated automatically.

Seven criteria have been removed in compliance with : PM1, PM3, PP2, PP5, BP1, BP3, BP6.

Review and edit the ACMG Classification of a CNV

The ACMG CNV Classification wizard is located in the Evidence tab of the Variant page. Itis available for tagged genomic variants.

The ACMG CNV Classification wizard is located in the Evidence section of the Variant page. It is available for tagged genomic variants.

The tool automatically scores sections 1, 2, 3, and partially scores sections 4 and 5 of the ACMG/Clingen guidelines, including the full PVS1 calculation required for intragenic variants. All the relevant data is summarized in an accessible table.

This tool is designed to save significant review time, reducing manual effort by up to 75–90% (ASHG 2020 abstract).

Factors considered in CNV scoring

When determining the automated classification, the system considers:

• Inheritance patterns – whether the CNV is de novo or segregates in a family

• Gene content – the number and type of genes affected, including ClinGen dosage sensitivity and predicted haploinsufficient genes

• Overlap with known pathogenic regions – alignment with established genomic regions linked to disease

By combining these elements, the scoring logic provides a clearer and more consistent starting point for CNV interpretation.

How do I review and edit the ACMG Classification of a CNV?

The ACMG SNV classification wizard features:

1. Automatically calculated ACMG class and score

1. ACMG score slider that shows the ranges of ACMG values for each classification and highlights where the current CNV falls:

   • Benign: ≤ –0.99

   • Likely Benign: –0.98…–0.90

1. Reclassify button that enables Edit mode

1. Gene Number:

   • Gene Number shows the total protein-coding RefSeq genes overlapped by the CNV. Of these:

     • Established ClinGen genes (dosage sensitivity or insensitivity defined by ClinGen scores)

1. Gene table that provides a summary of the affected protein-coding genes:

   • Gene description:

     • Name - HGNC gene symbol,

*Criteria with variable score:

• 2F, 2I

• 4A, 4B, 4C, 4D, 4E, 4I, 4J, 4K, 4L, 4M, 4N, 4O

• 5A, 5B, 5C, 5E, 5G, 5H

The Summary tab highlights core variant-related information from other tabs:

• (, , , ),

• (),

The SNVs classification engine applies the joint American College of Medical Genetics and Genomics/Association for Molecular Pathology (ACMG/AMP) guidelines for interpreting sequence variants using 28 evidence criteria (PMID: 25741868).

Emedgene independently determines the overall ACMG variant pathogenicity and .

ACMG variant classification

Emedgene applies the ACMG classification guidelines for variant interpretation using the standardized framework established by the American College of Medical Genetics and Genomics (ACMG). This approach is further refined with recommendations from the ClinGen Sequence Variant Interpretation (SVI) Working Group and other expert-reviewed publications to ensure consistent, evidence-based classification of genomic variants.

The Quality tab gives you a clear, interactive overview of variant quality across all sequenced individuals in a case. It displays zygosity, quality grades, and metrics tailored to each variant type—helping you evaluate data reliability and prepare for accurate interpretation.

Click on a family member’s icon to switch the sample in view.

Tips

Proband variant quality and caller

The overall variant quality based on variant caller-specific rules in a proband sample is highlighted on top of the Quality tab, next to the variant caller notation.

Pedigree card

In the pedigree card located on the left of the Quality tab, each sample is represented by zygosity and variant quality grade.

Quality metrics per variant type

A variant is assessed using a set of quality metrics tailored to its specific type, which are displayed next to the pedigree card.

SNVs/MNVs(100.39.0+)/Indels

• VCF Filter (DRAGEN 4.3+) — The FILTER column value in VCF indicating whether a variant passed the caller's quality thresholds or failed some checks.

• Base Quality (BQ) — Average Phred score for bases supporting the variant. High BQ indicates more reliable base calling.

• Depth (DP) — Total number of reads covering the position. It is the count of aligned reads after removal of duplicate reads and/or reads with MQ=0.

CNVs called by NGS pipeline

• VCF Filter (DRAGEN 4.3+) — The FILTER column value in VCF indicating whether a variant passed the caller's quality thresholds or failed some checks.

• Norm. depth of cov. (DRAGEN 4.2+, v100.39.0) — Normalized depth of coverage calculated by the .

• Copy Number — Estimated number of copies in the region.

CNVs and LOH/ROH called by array pipeline (DRAGEN Array 1.3)

• VCF Filter — Indicates if the variant passed DRAGEN quality checks.

• Likelihood Ratio — Measures confidence in the CNV call — higher values mean stronger evidence.

• Mean LRR (Log R Ratio):

Structural variants (SVs)

SV callers use breakpoint clusters, depth shifts, and split-read evidence to compute QUAL. Large SVs sometimes fail quality thresholds despite real biology being responsible.

• VCF Filter (DRAGEN 4.3+) — The FILTER column value in VCF indicating whether a variant passed the caller's quality thresholds or failed some checks.

• Quality

• Size

STRs

• VCF Filter (DRAGEN 4.3+) — The FILTER column value in VCF indicating whether a variant passed the caller's quality thresholds or failed some checks.

• Depth — Coverage at repeat locus.

• Repeat Number — Count of repeat units.

Haplotype variants

• Zygosity — Indicates whether the haplotype is present on one chromosome (heterozygous) or both (homozygous), based on the genotype reported in the VCF.

• VCF Filter (DRAGEN 4.3+):

  • Displays the value of the FILTER field from the VCF.

Allele distribution

Allele distribution card provides insight into how sequencing reads support reference and alternate allele for SNVs/MNVs/indels (including mtDNA), and STRs.

SNVs/MNVS/Indels

STRs

The In silico predictions card highlights aggregated scores for missense prediction, conservation, and splicing prediction. These are algorithmic assessments of variant effects based on known biological features, protein structures, evolutionary conservation, or machine learning models trained on large-scale data. The scores are calculated by proprietary algorithms that integrate outputs from individual in silico variant pathogenicity predictors.

These scores support , particularly PVS1/PP3 (pathogenic evidence) and BP4/BP7 (benign evidence), and are especially useful when experimental data is lacking.

The Evidence Tab provides tools to review, document, and classify variants efficiently. It combines manual and automated features to ensure accurate pathogenicity assessment, interpretation notes, and evidence visualization.

Users can import past interpretations, apply ACMG classification wizards, and generate evidence graphs—all within one workspace. Access is available only for tagged variants, ensuring that evidence review is focused and traceable.

1. Pathogenicity selection

A dropdown menu lets you assign the variant’s pathogenicity

Navigate between variants using the left and right arrow keys on your keyboard, or click the arrows on either side of the Variant page.

The Variant page showcasing the comprehensive variant information is accessible from the Variant table by selecting the corresponding variant row with a click.

Navigate between variants using the left and right arrow keys on your keyboard, or click the arrows on either side of the Variant page.​

The Variant page is comprised of:

1. . Displays Case ID, gene symbol, genomic DNA-level description of the variant, variant , and a link to your database. If the variant is already in your Curate database, you will see an Open Curate button. Otherwise, you will see an Export to Curate button.

2. Navigation panel (left).

3. Page body:

   1. . Highlights core variant-related information from other tabs

   2. . Reports essential variant- and gene-level information and indicates gene-related diseases.

4. Expandable (right). Records variant-level user activities, such as a variant, adding comments or evidence notes, or editing the evidence graph. Variant activity panel pops up upon clicking the Activities button.

ACMG automation module

Emedgene have implemented a technical automated solution for most criteria based on our scientific advisors’ recommendations and feedback from top clinical customers. For each criterion, we elaborate on the logic employed and the associated underlying thresholds. In addition, we give the user the flexibility to change the weight of specific criteria based on his professional judgment as recommended by ACMG/AMP guidelines.

ACMG criteria are mainly evaluated automatically using rule-based logic, allowing users to review and adjust evidence as needed. However, BS2 (partly) and PS4 need manual evaluation.

To learn how each criterion is assessed, select the corresponding evidence code in the table below.

^{_{Richards, S., Aziz, N., Bale, S., Bick, D., Das, S., Gastier-Foster, J., ... & ACMG Laboratory Quality Assurance Committee. (2015). Standards and guidelines for the interpretation of sequence variants: a joint consensus recommendation of the American College of Medical Genetics and Genomics and the Association for Molecular Pathology. Genetics in Medicine, 17(5), 405-424. doi:10.1038/gim.2015.30}}

Population Data

This section explains how population frequency data from large databases like gnomAD and ExAC is used to assess whether a variant is too common to be disease-causing. These criteria are part of the ACMG guidelines and help determine if a variant is more likely to be benign or pathogenic.

PS4 - Increased prevalence in affected individuals (Pathogenic, Strong-Supporting)

PS4 is applied when a variant is significantly more common in affected individuals than in unaffected controls, based on statistical comparison. This suggests the variant may play a role in causing the condition.

To apply PS4:

• There must be case-control data showing that the variant occurs more frequently in individuals with the disease.

• The strength of PS4 (Supporting, Moderate, or Strong) depends on the size of the study, quality of the evidence, and statistical significance (e.g., odds ratio, p-value).

This criterion is especially useful when working with common variants in complex disorders or when evaluating variants studied in large population cohorts.

PM2 - Rare or absent in population (Pathogenic, Supporting)

PM2 is applied when a variant is either very rare or completely absent in large population databases. This supports the idea that the variant might be disease-causing, because harmful variants are usually not common in healthy people.

In Emedgene, PM2 is automatically applied based on the gene and inheritance mode:

• For dominant conditions: the variant must occur in less than 0.01% of the population.

• For recessive conditions: the threshold is less than 1%.

• If there's known data in ClinVar for that gene, the system uses the highest known frequency of a confirmed disease-causing variant—up to a 1% cap—to fine-tune the cutoff.

This means:

• If a variant’s frequency is below the expected level for the disease—and particularly for that gene—PM2 will be applied.

• If a known pathogenic variant in the same gene has a high frequency, PM2 won’t be triggered unless the current variant is rarer than that.

This ensures that PM2 is more accurate and avoids flagging common variants as pathogenic.

BA1 - Very common variants (Benign, Strong)

BA1 is used when a variant is very common—specifically, if it appears in more than 5% of the population in large public datasets like gnomAD or in an internal database of at least 1,000 individuals. This level of frequency is too high for a variant to be responsible for rare genetic diseases, so BA1 provides strong evidence that the variant is benign.

BS1 - More common than expected (Benign, Strong)

BS1 is used when a variant is more common than you'd expect for a specific disease, even if it doesn’t reach the 5% threshold needed for BA1. It’s a strong indicator that the variant is likely benign.

In Emedgene, this is automatically calculated if the variant’s allele frequency is above a certain threshold that's specific to the gene and condition. These thresholds are calculated based on data from ClinVar, particularly variants that are pathogenic or likely pathogenic (P/LP) and have at least one-star review (meaning they’re well-reviewed and reliable).

The threshold also considers the mode of inheritance:

• For autosomal dominant (AD) diseases, the expected frequency is lower.

• For autosomal recessive (AR) diseases, a higher frequency might be acceptable.

This gene-specific logic improves how BS1 is applied and helps prevent mistakes in variant interpretation, especially for genes that naturally have more variation in the general population.

BS2 - Found in healthy people (Benign, Strong) (v100.39.0+)

BS2 is used when the variant is found in people who are healthy, even though the disease it’s associated with would normally appear in early childhood and be fully penetrant (i.e., definitely cause symptoms). If someone has the variant but is clearly unaffected, that’s a strong sign that the variant is not disease-causing.

BS2 is partially automated. BS2 is positively activated when the following all two conditions are met:

• Variant’s segregation pattern matches a known inheritance mode (AD, AR, or XLR).

• The variant is observed in public databases according to the expected zygosity

• Additionally, two conditions are manually set according to the observed state of the disease:

Computational and Predictive Data

Computational evidence plays a key role in classifying variants, especially when experimental data is unavailable. This section summarizes how Emedgene uses in silico (computer-based) predictions to support or refute a variant’s potential pathogenicity, in line with ACMG guidelines.

PVS1 - Predicted loss of function (Pathogenic, Very Strong)

PVS1 is applied when a variant is predicted to result in a loss of function (LoF) in a gene where such a mechanism is known to cause disease. This includes variants such as nonsense (stop-gain), frameshift, canonical splice site changes, or large deletions that disrupt essential exons. Emedgene uses a structured, evidence-based decision tree aligned with the ClinGen framework (Abou Tayoun et al., 2018) and transcript-based considerations outlined by Walker et al. (2023).

To determine the appropriate evidence strength (Very Strong, Strong, Moderate, Supporting), the platform evaluates multiple factors:

• Whether the affected exon is biologically important (e.g., present in canonical or disease-relevant transcripts).

• Whether the variant is likely to trigger nonsense-mediated decay (NMD).

• Whether the disrupted region is known to be critical for protein function.

A graphical interface visually traces this decision logic in the Emedgene platform, enabling users to see exactly how the PVS1 tag was determined and what supporting evidence was used.

References:

• Abou Tayoun et al., 2018 – ClinGen LoF Framework

• Walker et al., 2023 – Transcript and functional considerations for LoF variants

PS1 - Same amino acid change as known pathogenic variant (Pathogenic, Strong) (v100.39.0+)

PS1 is used when a variant results in the same amino acid change as a previously confirmed pathogenic variant, but via a different nucleotide change.

This implies that the functional consequence is identical, even if the underlying DNA sequence is different. For example, two different codons could both result in the same substitution at the protein level.

The defined logic supports both missense and splice variants. For missense, PS1 is applied when the variant leads to the same amino acid substitution as a ClinVar pathogenic variant on the same codon. For splice variants, PS1 is applied when the variant affects the same splice site or region as a ClicVar pathogenic variant, with similar predicted impact based on SpliceAI and SpliceAI-10K.

Strength is assigned based on variant type, location, and ClinVar classification. Supporting evidence includes links to the ClinVar variant, SpliceAI score, and predicted effect.

References:

• Walker et al., 2023 – Computational and RNA-splicing evidence / bioinformatic codes

PM5 - Novel missense at known pathogenic residue (Pathogenic, Moderate)

PM5 applies when a new missense variant occurs at a protein position where other different missense changes have already been classified as pathogenic.

This suggests that the position is critical for protein function and intolerant to change. Emedgene cross-references the variant against known pathogenic variants at the same amino acid site to apply this criterion.

PM4 - In-frame indels or stop-loss outside repeats (Pathogenic, Moderate)

PM4 is used when a variant leads to an in-frame deletion or insertion, or a stop-loss, outside of a repetitive region of the gene.

In-frame variants do not shift the reading frame, but they can still alter protein function if they affect a structurally or functionally important region. Emedgene checks whether the variant is located within low-complexity or repetitive regions before applying PM4, increasing the specificity of this evidence.

PP3 - Damaging computational predictions (Pathogenic, Supporting-Strong)

PP3 supports pathogenicity when multiple in silico prediction tools suggest that a variant is deleterious. This is typically applied to missense or splicing variants and can vary in strength based on the confidence of the tools. Emedgene uses thresholds aligned with Pejaver et al. (2022) and Walker et al. (2023).

In Emedgene, this includes:

• Missense impact (REVEL score)

• Splicing impact (SpliceAI)

• Conservation scores (evolutionary constraint)

The strength of PP3—Supporting, Moderate, or Strong—is determined by how confident the predictions are:

• REVEL score (Pejaver et al., 2022):

  • 0.644–0.773: Supporting

  • 0.773–0.932: Moderate

A variant must meet at least two supportive predictions across conservation, protein impact, or splicing tools to qualify. Tools may include REVEL, SpliceAI, dbscSNV, and conservation metrics like GERP or SiPhy.

This scoring ensures a standardized, evidence-based use of computational tools in variant interpretation.

References:

• Pejaver et al., 2022 – Benchmarking REVEL for clinical variant prediction

• Walker et al., 2023 – Standardizing splicing predictions using SpliceAI

BP1 - Missense variant in a truncation-driven gene (Benign, Supporting)

BP1 applies to missense variants found in genes typically associated with disease through loss-of-function mechanisms (e.g., nonsense or frameshift), not missense.

Emedgene evaluates ClinVar data to calculate a benign-to-pathogenic ratio, and applies BP1 when there is at least a 5:1 ratio in favor of benign variants at the same gene locus. This is supporting evidence for benignity, particularly when missense variants are less likely to be disease-causing in that gene.

BP4 - Benign computational predictions (Benign, Supporting-Very Strong)

BP4 supports a benign interpretation when multiple computational predictors suggest the variant is not likely to be damaging. As with PP3, the strength of BP4 can vary depending on the tool confidence.

For missense variants, REVEL thresholds are:

• Very Strong Benign: < 0.003

• Strong: 0.003–0.016

• Moderate: 0.016–0.183

For splicing, BP4 applies when SpliceAI score ≤ 0.1, suggesting minimal or no disruption to splicing.

As with PP3, a minimum of two independent predictors must agree for BP4 to be applied. This ensures evidence consistency across protein function and splicing tools. Emedgene uses these thresholds to adjust the strength of BP4 accordingly, making the evaluation more consistent and aligned with recent research.

References:

• Pejaver et al., 2022 – REVEL thresholds

• Walker et al., 2023 – SpliceAI benign cutoffs

BP7 - Synonymous variant with low impact (Benign, Supporting)

BP7 is applied to synonymous (silent) or non-coding intronic variants that are not predicted to affect splicing and fall in non-conserved regions. Emedgene applies BP7 only when the following are true:

• The variant is synonymous or deep intronic

• SpliceAI score is ≤ 0.1, confirming no splicing disruption

This criterion supports benign interpretation and helps users confidently classify silent variants that otherwise appear suspicious due to location in exons or introns.

Reference:

• Walker et al., 2023 – Updated guidance for BP7 using splicing prediction

Functional Data

Functional data criteria assess the results of experimental studies—both in vitro (in the lab) and in vivo (in living systems)—to determine whether a variant disrupts gene or protein function. This type of evidence is powerful because it reflects direct biological testing rather than predictions alone.

PS3 - Functional studies show damaging effect (Pathogenic, Variable Strength)

PS3 is applied when reliable experimental studies demonstrate that a variant has a deleterious effect on the gene or its protein product. These effects might include:

• Loss of enzymatic activity

• Disruption of protein structure or folding

• Impaired molecular interactions

The studies used must be well-validated, reproducible, and performed in a relevant biological context. Emedgene surfaces supporting publications automatically by scanning literature databases for functional assays relevant to the variant under review, as outlined in Walker et al., 2023 and Brnich et al., 2020.

The strength of PS3 (Supporting, Moderate, or Strong) depends on:

• The type and quality of the assay

• The number of independent studies showing consistent results

• How closely the assay reflects the true disease mechanism

For example, a high-quality cell-based assay showing complete loss of protein function in a well-characterized disease gene could support PS3_Strong, while a partial defect in a less direct assay might be PS3_Moderate. All evidence is traceable in Emedgene, and users can adjust final strength based on their own review.

References:

• Walker, C.E. et al., 2023 – Updated guidance for integrating functional and computational evidence into variant classification.

• Brnich, S.E. et al., 2020 – Recommendations for the application of functional evidence in clinical variant interpretation (American Journal of Human Genetics).

PM1 - Missense variant in mutational hotspot or critical functional domain (Pathogenic, Moderate)

PM1 is assigned when a missense variant falls within a mutational hotspot or critical protein domain that:

• Has > 70% pathogenic missense variants reported

• Has at least 10 ClinVar entries

• Shows no benign variation in population databases

These regions often correspond to active sites, binding domains, or structural motifs critical for protein function. Variants in these areas are more likely to have clinical impact. PM1 is not applied to mtDNA variants.

PP2 - Missense variant in gene where missense is a commom mechanism (Pathogenic, Supporting)

PP2 applies to missense variants in genes where:

• Disease is frequently caused by missense changes (often affecting critical functional regions)

• Benign missense variation is rare

Emedgene calculates a pathogenic-to-benign ratio using ClinVar data, applying PP2 when this ratio is ≥ 2:1 with at least 10 total entries for the gene. This ensures statistical reliability. PP2 is not applied to mitochondrial DNA (mtDNA)variants.

In practice, this means if a gene is known to cause disease through subtle amino acid changes rather than truncating mutations, and nearly all missense changes are pathogenic, a new missense variant is more likely to be disease-causing.

BS3 - Functional studies show no damaging effect (Benign, Variable Strength)

BS3 is the benign counterpart to PS3. It is applied when robust functional evidence demonstrates that a variant does not impair the gene’s or protein’s function. This can include:

• Normal activity levels in enzyme assays

• Proper localization in cell-based imaging

• Correct protein folding and stability

As with PS3, Emedgene’s automated literature classifier identifies relevant publications and integrates them directly into the variant’s evidence record, helping users avoid manual searches.

The strength of BS3 depends on:

• The assay’s relevance to the known disease mechanism

• The number of independent lines of evidence showing normal function

• Agreement with other evidence (e.g., population frequency)

A high-quality study showing full wild-type function in multiple models may support BS3_Strong, whereas evidence from a single model may support BS3_Supporting.

References:

• Walker, C.E. et al., 2023 – Updated guidance for integrating functional and computational evidence into variant classification.

• Brnich, S.E. et al., 2020 – Recommendations for the application of functional evidence in clinical variant interpretation (American Journal of Human Genetics).

Segregation Data

PP1 - Cosegregation with disease in multiple affected family members (Pathogenic, Variable Strength) (v100.39.0+)

PP1 is applied when there is clear evidence that a variant segregates with the disease in a family. For this criterion to be met in Emedgene, all three of the following must be true:

1. At least two affected family members are included in the case data.

2. The variant co-occurs with the disease phenotype in the pedigree — meaning every affected person carries the variant.

3. The variant is in a gene that is definitively known to cause a disease that matches the observed phenotype in the family.

PP1 strength is determined according to the combination of affected and unaffected family members within the pedigree.

In practice, PP1 provides stronger evidence as the number of segregating family members increases. For example, if five affected relatives across multiple generations all carry the same variant in a gene known to cause the disease, this can significantly raise confidence in a pathogenic classification.

Reference:

• Richards et al., 2015 – ACMG Guidelines for the Interpretation of Sequence Variants; Emedgene Help Guide – Segregation Evidence Logic

• Biesecker et al., 2024 – ClinGen guidance for use of the PP1/BS4 co-segregation and PP4 phenotype specificity criteria for sequence variant pathogenicity classification

BS4 - Lack of segregation in affected members of a family (Benign, Strong) (v100.39.0+)

BS4 is applied when there is clear evidence against segregation, meaning that the variant does not track with the disease within a family. For BS4 to be assigned in Emedgene, all of the following conditions must be met:

1. The case is not a singleton and there must be more than one affected family member included.

2. The lack of segregation is consistent across available pedigree data and not explainable by incomplete penetrance or phenotypic misclassification.

When these conditions are met, BS4 can strongly support a benign classification, especially for conditions with high penetrance, where all affected individuals would be expected to have the causal variant.

Reference:

• Richards et al., 2015 – ACMG Guidelines for the Interpretation of Sequence Variants; Emedgene Help Guide – Segregation Evidence Logic

• Biesecker et al., 2024 – ClinGen guidance for use of the PP1/BS4 co-segregation and PP4 phenotype specificity criteria for sequence variant pathogenicity classification

De novo data

De novo evidence is applied when a genetic variant appears for the first time in an affected individual (proband) and is absent in both biological parents. This type of evidence can strongly support pathogenicity, especially for disorders known to occur from new mutations rather than inherited variants. Emedgene evaluates this automatically based on case data, genomic analysis, and phenotype matching.

PS2 - Confirmed de novo variant (Pathogenic, Strong to Moderate)

PS2 is assigned when a variant is confirmed to be de novo in a patient who has the associated disease, with both paternity and maternity genetically validated. To meet the PS2 criteria in Emedgene:

1. The proband must carry the variant in a heterozygous state (or hemizygous if male for X-linked variants).

2. Both parents must be wild-type (reference) for that position in their genotypes and clinically unaffected.

3. Parentage confirmation — both maternity and paternity — must be established through genetic testing in the lab module, ensuring the variant is truly new and not inherited.

Emedgene also integrates phenotype specificity scoring using the Phenomeld engine. This ensures that the de novo event is evaluated in the context of how well the patient’s clinical features match the disease linked to the gene. Based on ClinGen SVI Working Group (2021) recommendations:

• If the phenotype matches at the PP4 threshold, PS2 is applied at Strong strength.

• If the phenotype match score is ≥0.8, PS2 is applied at Moderate strength.

• If the phenotype match score is ≥0.4, PS2 is applied at Supporting strength.

This structured approach makes PS2 more consistent and reliable, reducing false positives from unrelated phenotypes.

References:

• Richards, S. et al., 2015 - Standards and Guidelines for the Interpretation of Sequence Variants: A Joint Consensus Recommendation of the ACMG and the AMP. Genetics in Medicine.

• ClinGen Sequence Variant Interpretation (SVI) Working Group, 2021 - Recommendations for De Novo Criteria (PS2 & PM6), Version 1.1.

• Phenomeld – Emedgene’s phenotype-matching system for clinical correlation in variant curation platform Emedgene Analyze.

PM6 - Assumed de novo variant without parentage confirmation (Pathogenic, Moderate to Supporting)

PM6 is used for cases where the variant appears to be de novo, but parentage is not genetically confirmed. The application conditions are the same as for PS2, with the following differences:

1. The proband is heterozygous (or hemizygous) for the variant.

2. Both parents appear wild-type for the variant and are clinically unaffected.

3. Parentage confirmation data is unavailable — either because it was not tested or not recorded.

Phenotype matching again plays a key role in determining PM6 strength:

• PP4 threshold phenotype match → PM6 at Moderate strength.

• Phenotype score ≥0.8 → PM6 at Supporting strength.

• Phenotype score ≥0.4 → PM6 at Supporting strength (lower confidence).

PM6 provides valuable evidence when parentage testing cannot be performed but should be interpreted cautiously, especially for disorders that might have late-onset or incomplete penetrance.

References:

• ClinGen Sequence Variant Interpretation (SVI) Working Group, 2021 - Recommendations for De Novo Criteria (PS2 & PM6), Version 1.1.

• Phenomeld – Emedgene’s phenotype-matching system for clinical correlation in variant curation platform Emedgene Analyze.