Skip to main content
All CollectionsCase Analysis Workbench
Franklin's AI-based variant prioritization engine
Franklin's AI-based variant prioritization engine

Learn more about Franklin's artificial intelligence-driven algorithms to pinpoint the relevant variants in case analysis.

Luciana De Cesare avatar
Written by Luciana De Cesare
Updated over a year ago

The Franklin platform has an advanced artificial intelligence-driven engine designed to prioritize and interpret variant data. This powerful engine extracts a range of features from diverse sources of evidence and identifies the most probable causal pathogenic variants to add them to the Workbench for the user's review. It compiles evidence from various channels, including public databases, literature-based sources, exclusive Franklin Community data, and in-house curations.

The engine offers insights and prioritization for different types of variants, including short variants (SNPs and indels), copy number variants (CNVs), structural variants, and compound variants, where one variant may be an SNP and the other a CNV.

To select the variants that are highlighted in the Workbench, the core elements on which the priority engine relies include:

- Variant classification

- Genotype association with clinical and phenotypical evidence

- Family segregation based on different inheritance models

- Known inheritance models for the gene/disease

- Level of technical confidence in the variant, as determined by the variant caller

Franklin's AI-based classification engine adheres to ACMG/AMP standards and guidelines. Additionally, it integrates supplementary knowledge from databases and literature to expedite gene-specific variant classification. The AI-based classification algorithms evaluate the classification criteria by incorporating gene-specific evidence from more than 100 different gene and variant classification sources, such as ClinVar, ClinGen, Uniprot, and gnomAD.

In addition, Franklin's genotype-to-phenotype and disease association engine consolidates and assesses relationships between genes, diseases, and phenotypes, drawing from an extensive range of databases, including Orphanet, Monarch, HPO, and DECIPHER. Moreover, it incorporates novel evidence by utilizing a text-mining engine that scans public literature.

The relationships between the patient's clinical terms and the genes are scored and ranked. The prioritization algorithm factors in the strength of the connections between the patient's phenotypes/disease and the gene.

For variant prioritization, the algorithm takes into consideration the possible inheritance models per variant, which can apply to single sample analysis, a proband of a trio, or a larger family pedigree. It integrates this information with the supported gene inheritance models derived from reputable published and curated sources, enabling evaluation of the consistency features of the inheritance models and generation of an overall score.

To address the potential presence of false variants resulting from sample contamination or sequencing artifacts, which can lead to erroneous pathogenic variant identification, the algorithm incorporates diverse quality metrics and confidence data, factoring in the probability of a variant being a false positive.

By synthesizing all the aforementioned features and evidence, the Genoox artificial intelligence-based prioritization engine excels at pinpointing the variants that are likely to be clinically relevant in the patient's context, adding them to the Workbench.

Still have questions? Reach out to our Support Team, they'll be happy to help!

Did this answer your question?