Franklin has a powerful filtering system that allows users to perform complex queries on genomic data. With multiple advanced filters at your disposal, you can easily sort through vast amounts of data and extract the information you need.
In this article, we'll walk you through the functionality of each filter, providing detailed explanations and guidance on how to use them effectively. Whether you're a seasoned researcher or a newcomer to the field, this guide will help you make the most of Franklin's filtering capabilities.
Note that you have the option to save multiple filters together as a “quick filter” - more details to be found in this article - “Managing quick filter”
Singleton - SNV and indels filters:
Phenotypes - the selected phenotypes will influence Franklin's sorting when using the "by priority" option. This will display variants in the genes associated with these phenotypes at the top. Toggling the "use as filter" option will filter out variants in genes that are not associated with the selected phenotypes.
Gene properties:
Gene inheritance - Filter the known inheritance patterns of genes based on their associated conditions. Franklin uses multiple sources, such as OMIM, MONDO, Orphanet, Decipher, GENCC, and others, to build its Gene-Disease associations.
Mechanism of disease:
Sensitivity to LOF - filter based on the gene constraint metric of observed/expecated ratio in gnomAD. Genes are shown if pLoF o/e upper score <= 0.35, or if their pLI score >0.9. To read more on gnomAD 'Gene constraint' metrics - visit https://gnomad.broadinstitute.org/help/constraint.
Sensitivity to missense - filter based on the gene constraint metric of observed/expecated ratio in gnomAD. Genes are shown if missense o/e upper score<=0.35. To read more on gnomAD 'Gene constraint' metrics - visit https://gnomad.broadinstitute.org/help/constraint.
Public curated panels - Allow filtering based on known public gene panels, such as ACMG secondary findings or OMIM morbid genes.
Franklin's Classification - This feature filters variants according to Franklin's automated classification. Note that the VUS classification is divided into three classes: VUS Leaning Benign, VUS, and VUS Leaning Pathogenic.
My organization's classifications - Filter your search results according to your own curated classifications saved in the organization's knowledge base. Note that for a given variant, there may be different classifications for different diseases. For each disease, Franklin will use the latest available classification.
Compare with - This filter is used to compare variants with other singleton samples in your cases list. To use the filter, type in the sample name you want to compare with, and then select additional parameters - shared/unique and identical/opposite/any. You can add multiple samples to the comparison by selecting them one-by-one.
Shared variants look for variants shared between the samples.
Identical - Variants have the same zygosity.
Opposite - Variants have opposite zygosity.
Unique variants look at variants that exist in the sample but not in the compared sample.
Identical - Includes all unique variants and shared variants but with a different zygosity in the compared sample.
Opposite - Includes all unique variants and shared variants with the same zygosity.
Panels - you can add gene panels from the organization's knowledge base to filter the results and only show variants in the genes within the selected panels. You can apply multiple panels together. If a panel is saved as part of a quick filter, it will appear as a "quick filter panel" instead of a case panel. The case panels are also displayed in the panels section of the workbench.
Variant type - filter results based on the variant type, either SNV or indel.
Region - filter results based on the region of the variants, such as exonic, intronic, or UTR 3', etc.
Effect - filter results based on the effect of the variants, such as synonymous, missense, or frameshift, etc.
Regulatory - Filter based on Vista and Ensembl regulatory build, including both enhancers and promotors.
Chromosome - filter based on genomic location. Select a chromosome and a start/end position.
Zygosity:
Filter variants based on zygosity: Homozygous, Heterozygous, Homozygous Reference (wild type), and Uncalled Zygosity.
Regions of Homozygosity (ROH): Variants that are within regions of homozygosity. The list of ROH is displayed in the Workbench next to the QC section.
Suspected Compound: This filter shows variants that might be in compound with another variant in the same gene. The filter searches for a compound pair variant that is possible relevant in terms of classification/region/effect/confidence (classified by Franklin as VUS, VUS-leaning pathogenic, likely pathogenic, or pathogenic AND region is exonic/splice AND effect is not synonymous AND confidence is Med/High).
The number of possible compound variants is displayed in the expanded variant tile and fully displayed in the variant full details view under the "variant assessment" tab.
Suspected Compound SV: This filter shows variants that might be in compound with another CNV variant that is overlapping the same gene. The compound CNV is classified as either VUS, likely pathogenic, or pathogenic.
Allele Balance (VAF) - filter variants based on their allele balance. The allele balance is displayed on the tile, under the confidence value, and also in the expanded tile under the confidence section. You can either select one of the predefined presets or use the pencil icon to manually select the range.
Frequency - This section provides options to filter variants based on multiple population databases and internal frequency. You can either select one of the predefined presets or use the pencil icon to manually select the range.
Internal Frequency - filter variants based on their frequency within the organization's case cohort. The internal frequency is updated once a day. You also have the option to filter based on the number of cases in which this variant appeared, which can be helpful when the case cohort is still small.
Aggregated Frequency - this value is the aggregation of several databases, but places a high weight on the gnomAD database.
Confidence: This filter allows you to filter variants based on their technical confidence. The confidence value is calculated using multiple metrics such as quality, depth, strand bias, and others. The calculation method differs between different variant callers. Additionally, you can filter by raw data from the variant caller, such as variant quality, quality by depth, and depth.
Predictions: Filter variants according to the values of multiple in-silico prediction tools. In addition, an aggregated prediction filter is available, aggregating multiple in-silico tools according to the variant effect/region. Read more on Fraknlin's aggregated prediction calculation here.
ClinVar evidence: Filter variants according to ClinVar submission classifications values and number of submissions.
Variant callers: Filter variants according to the variant caller that called the variant. This supports Franklin's approach of using multiple variant callers. Note that you can switch between 'and', 'or', and 'only'.
Singleton - additional SV filters:
Gene Properties - Mechanisms of Disease
Haploinsufficient - based on ClinGen dosage sensitivity curation. This will show genes that are curated to be haploinsufficient with a score of 1-3. To read more on ClinGen dosage sensitivity visit https://clinicalgenome.org/curation-activities/dosage-sensitivity/
Triplosensitive - based on ClinGen dosage sensitivity curation. This will show genes that are curated to be triplosensitive with a score of 1-3. To read more on ClinGen dosage sensitivity visit https://clinicalgenome.org/curation-activities/dosage-sensitivity/
Estimated Copy Number - Filter according to the estimated copy number of the variant. You can either select one of the pre-defined presets or use the pencil icon to apply a manual range.
Variant Type
Filter by different types of variants such as deletion, duplication, and repeat expansions.
Filter by variant length.
Are there mosaic variants?
HaploInsufficient -
Triplosensetive -
Occurrences - This filter is used to select variants based on their frequency in normal population databases, such as gnomAD, DGV, and DGV gold. The occurrence annotation counts variants that have at least 50% reciprocal overlap with the variant being analyzed. The aggregated occurrences is simply the sum of all the databases together.
Internal occurrences - This is similar to the occurrences annotation described above, but is based on the cohort of internal organization cases.
Family-specific filters:
Inheritance - This filter is intended to filter on the calculated inheritance pattern of each variant within the family analysis. Note that this is different than the gene inheritance filter in the ‘gene properties’ filter group.
Single inheritance:
Autosomal recessive - The affected individual/s carries a variant on both alleles (homozygous), while all unaffected individuals are heterozygotes and thus only carriers. For example, in the classic trio analysis, both unaffected parents are carriers (heterozygotes) of the variant, and the affected proband is homozygous for the variant.
Autosomal dominant - The affected individual/s carries a variant on a single allele (heterozygote). The other unaffected individuals will show the reference wild-type genotype. For instance, when both the proband and the affected parent are affected, the other parent will have the wild-type genotype.
X Linked Recessive - A variant in a gene on chromosome X causes the phenotype to be expressed in males, while females must carry the mutation on both the X chromosomes in order to be affected.
X Linked Dominant - A variant in a gene on chromosome X causes the disease heterozygously, while only one copy of the alleles is sufficient when inherited from a parent who has the disorder.
Y Linked - A pathogenic variant in a gene on chromosome Y causes the disease and is only presented in males.
De Novo - A variant that is present only in the proband, i.e. not present in either parent. Note that there might be results in this model that are actually sequencing errors and not necessarily a de novo mutation.
Compound Heterozygote - The affected individual/s carries two different heterozygous mutations on the same gene. Each mutation is either inherited heterozygously from each healthy parent or is de novo.
Compound SNP/SV - The affected individual/s carries two different heterozygous mutations on the same gene. Each mutation is either inherited heterozygously from each healthy parent or is de novo. One of the variants is a CNV and the other is an SNV or indel.
Model Strictness: Model Strictness controls the strictness or leniency of the inheritance model. In dominant inheritance models, the most strict model would be similar to "complete penetrance," while the most lenient would be similar to "low penetrance."
In all models, the "lenient" option would include more variants, some of which may not completely match the inheritance model due to sequencing errors or low quality, among other factors.Moving the radio button to the middle state allows for a single family member to not comply with the selected inheritance pattern. For example, when searching for dominant inheritance and both parents are unaffected, zero variants will be displayed. Toggling the radio button to the center location will also present variants that are heterozygous in both the affected individual (proband) and a single unaffected parent, which represents a situation of partial dominant inheritance pattern.
Toggling the radio button to the rightmost option allows for two family members to not comply with the selected inheritance pattern.
Family zygosity - filter according to all family members’ zygosity.
Still have questions? Reach out to our Support Team, they'll be happy to help!