Algorithmic filters

Algorithmic filters are very powerful filters allowing you to perform more complex variant filtering. These filters can be fully customized by our experts according to the specific needs of your workflow. Use cases for algorithmic filters are almost infinite. 

As an example, a custom-made algorithmic filter is developed to:

1. Keep any variants that fall either within or -50/+20 from the edges (except for the first and last exon) of coding exons of genes in a gene list, but only if they are within the transcript: exclude variants that are 50nt upstream of the first exon or 10 nt downstream of the last.
2. Exclude variants that have variant fraction lower than 0.15.
3. Exclude variants if they fall in positions with a coverage of less than 20.
4. Keep only variants which match any of these conditions:
   • marked as Pathogenic or Likely Pathogenic by VarSome's Germline Variant Classification OR
   • marked as "Not provided" or "Uncertain Significance" or "Likely Pathogenic" or "Pathogenic" or "Drug Response" or "Drug Association" or "Risk Factor" or "Conflicting" or "Cancer" or "NA" or "Other" in ClinVar

Below you may find some examples of algorithmic filters that are available on VarSome Clinical:

• Recessive candidates
• ACMG Actionable Gene
• Carrier risk for couples (n= 2)
• Compound Heterozygous Candidates (n= 1)
• De novo (strict) (n= 3)
• De novo candidates (naive) (n= 3)
• Segregating Variants (n≥ 2)
• Variants in Common (n≥ 2)
• Fisher exact test (n≥ 4 )
• VarSome Picks

Once on the page with the list of all Algorithmic filters, you can always click on the blue circle to learn about the details of each Algorithmic filter (for CLG Carrier Genes, in the example below):

You can also find demonstrated examples of how to apply them in your analyses.

You can access Algorithmic filters under Launch analysis > Algorithmic filter analysis:

For several of the existing filters since release 11.3, we provide the option to change specific parameters that used to be fixed. That way you can customize each of those filters according to your needs. For some of the filters, in order to add the parameterization functionality we have integrated similarly functioning filters into one. For example, you can now find under the filter “Segregating Variants”, the filters are previously known as: 

• Segregating Variants (dominant, all VUS), 
• Segregating Variants (dominant, strong VUS)
•  Segregating Variants (recessive, all VUS)
•  Segregating Variants (recessive, strong VUS)
•  Compound Heterozygous Segregating Variants (all VUS)
•  Compound Heterozygous Segregating Variants (strong VUS).

You will see next to certain filters a light blue “Options” button:

After you choose from the drop-down the analysis to which you wish to apply one or more filters, you then need to select the filters by clicking the relevant check-boxes, which will in turn activate the options box.

This box displays the default parameters that we use for this particular dynamic filter.

Once the analysis has run, you can view the filter options selected for the algorithmic filter analysis both in the sub-analysis name

and in the “Sample/Analysis Information” option of the “Analysis Actions” menu:

Exonic & Splicing algorithmic filter

A special case of algorithmic filter is “Exonic and splicing” which only keeps exonic (including UTR and other non-coding exons) and splicing (no more than 10 nucleotides from a known splice site) variants. 

The filter may be launched manually, just like any other, but it will also be run automatically on any analyses with more than 500,000 variants. 

This algorithmic filter provides the same results as would occur if you made a dynamic filter with the following criteria:

• Coding
• Splicing
• Non-coding exon +3’ UTR
• Non-coding exon +5’ UTR 

The aim of this filter, and the reason it will run automatically for large analyses, is to provide a smaller subset of results to the user which will be far quicker and easier to sort through. Since, even with WGS analyses, the variants of interest tend to be those that can affect the protein sequence, we feel that this filter will help our users quickly identify and focus on the variants of interest even on larger samples such as WGS.

Later on, you lay over Dynamic filters to fine-tune the list of your variants.