Inspired by the ClinGen variant curation expert panels (VCEPs), Franklin implements the frequency rules in a gene-specific manner. This means that specific thresholds are assigned to each gene.
Franklin learns the frequency of the gene, by looking at all of the known pathogenic or likely pathogenic variants in the genes, and looks at their gnomAD frequencies across the different subpopulations. These variants include variants that were reported in ClinVar of Franklin Community, after removing ones that are likely to be false classifications.
Similar to the ClinGen VCEPs, based on the most common variants in the gene, it assigns the relevant thresholds.
In addition, Franklin supply the relevant information in case one wants to review it himself, as well as to get more information about the gene:
A significant advantage of this approach is robustness and consistency. While some VCEPs calculate their thresholds using the overall gnomAD frequency, others use the gnomAD sub-populations frequencies.
In addition, thresholds could vary depending on which gnomAD version is being used. This means that thresholds which were being recommended based on a specific gnomAD version (and a reference genome version), are likely to be different when the gnomAD database is being updated.
Franklin is constantly being updated and the classification and thresholds are synced with the updated and relevant data.
While this significantly helps to remove uncertainty and correctly classify variants as benign or likely-benign, Franklin takes cautious steps to make sure not to assign too low of a threshold.
There must be enough information (enough classified P/LP variants) about the gene in order to assign gene-specific thresholds.
A minimum threshold of 0.05% is applied, even if the most frequent variant is below that.
As recommended by ClinGen and the ACMG, caution is being considered in the case of founder-population (Finnish, Ashkenazi Jewish), which requires at least 1% in order to apply a benign rule. So for example, if the threshold of a gene is 0.3% for applying supportive rule BS1, in case only the Finnish population is above this threshold, it would be applied only in the case of more than a 1% frequency.
In some genes, there might be outlier variants with a relatively very high frequency possibly due to a founder effect, that might inflate the frequency. In line with the SVI recommendations, Franklin automatically detects these outlier variants and would not apply the BA1 rule for them, as well as would keep the correct recommended thresholds. For example, GJB2:c.109G>A or G6PD:c.563C>T are listed under their gene outlier variants list.
The outlier pathogenic variants in a gene are based on a statistical analysis we perform among the pathogenic variants in each gene. This information is constantly being updated based on new evidence.
Additional information can be found in the poster we presented at the 2018 ACMG conference.
Still have questions? Reach out to our Support Team, they'll be happy to help!