SOPHiA GENETICS™ is excited to be a part of the American Society of Human Genetics (ASHG) Annual Meeting 2023, held in person in Washington D.C., USA, from November 1-5.
You will have the chance to chat with our experts at booth #1923 and demo the universal SOPHiA DDM™ Platform.
From Concealed to Revealed: Advancing NGS variant interpretation
Friday, November 3
3:30 – 4:30pm
The critical role of transcript analysis for refining the classification of variants associated with constitutional disorders
Kai Lee Yap, PhD
FACMG, Director of Molecular Diagnostics, Ann & Robert H. Lurie Children’s Hospital of Chicago
Assistant Professor of Pathology, Northwestern University Feinberg School of Medicine
Pinpointing new pathogenetic variants associated with rare disorders in the Brazilian population
Diego Miguel, MD, MSc
Hereditary Cancer Geneticist, Clinica AMO, Salvador, Brazil
Thursday, Nov 2 & Friday, Nov 3 – 3-5 pm
Saturday, Nov 3 – 2:15-4:15 pm
Statistically robust familial variant analysis
First Author: Adrian Janiszewski
Familial variant analysis (FVA) can be a powerful tool for the identification of disease-causing germline variants. FVA investigates variant segregation patterns and excludes variants that do not follow the inheritance patterns of interest, reducing the number of candidate disease-causing variants that need to be further assessed. However, the sequencing noise present in large-panel sequencing may cause errors in variant zygosity estimation, hindering the filtering ability of FVA. This effect is expected to grow with increasing pedigree size, as sequencing errors in one individual can negatively affect the analysis of the entire family.
Here, we develop an FVA algorithm that combines sequencing data and affection status of family members in a statistically robust way and guarantees full control over the false negative rate. Our approach is based on characterizing the consistency of sequencing data with the expected variant zygosity, in each family member, for all inheritance modes available, and representing it as a set of p-values with the subsequent combination of the relevant scores using Fishers p-value combination procedure. This approach supports the analysis of pedigrees of any size, and data missingness: it leverages on affection status of a family member even if sequencing data is unavailable, extending the use of FVA in clinical sample analysis.
Using in-silico simulations of family data, we demonstrate the controlled false negative rate of our algorithm and estimate its efficiency defined as the fraction of variants that do not follow the tested segregation pattern that are correctly filtered out. The experiments cover autosomal recessive, autosomal dominant, X-linked and compound heterozygous variants propagating in pedigrees with up to 3 generations and siblings. For each case, we simulated 100 families with average read coverage of 30 and 100 and defined noise model. We show that the filtering efficiency increases with increasing availability of affection status and/or sequencing data of individuals. For example, filtering power for autosomal dominant variants increased by 16,2% if grandparents’ affection status was added to a trio analysis, and by 45% if grandparents’ sequencing data were also added.
Altogether, our FVA algorithm provides enhanced variant filtering capabilities for identification of disease-causing variants from large NGS data such as whole-exome or whole-genome sequencing. Combined with the ability to analyse pedigrees of any size and to include individuals for which only affection status is available, our algorithm provides means to improve the efficiency and accuracy of variant analysis and interpretation.