2023 ACMG Annual Clinical Genetics Meeting Digital Edition: Poster Gallery-Developing Bayesian graphical models to provide continuous, probabilistic variant interpretation

Pdf Summary

The document discusses the allele frequency model and the importance of variant interpretation in inferring the relationship between genetic variants and diseases. Traditionally, variants are classified into categories, but this leaves a large number of variants as uncertain. The document proposes using probabilistic graphical models (PGMs) to incorporate population allele frequency and clinical phenotype data in variant interpretation. PGMs are flexible Bayesian statistical frameworks that leverage domain expertise to perform probabilistic inference. The probability that a variant is pathogenic can be predicted using this approach. The document also mentions the Genome Aggregation Database (gnomAD) and Invitae's own research-eligible sample database as valuable sources of data for variant interpretation.<br /><br />The document explains the data generative process for allele counts and the use of additional features like constraint scores to enhance the model's accuracy. Pathogenicity likelihood can be estimated for new alleles with high accuracy. The document also includes graphs and examples to illustrate the interpretation process.<br /><br />The document emphasizes that probabilistic graphical modeling provides a suitable framework for modeling the complex relationships in variant interpretation. Domain-specific models like the allele frequency model and the clinical phenotype model can be combined into an overall variant interpretation model. This approach allows for more granular variant interpretations and the precise probability of pathogenicity can be reported instead of relying on the current five-category framework. The use of condition phenotypes learned from affected patients further enhances the prediction of variant pathogenicity.<br /><br />Overall, the document highlights the importance of probabilistic graphical modeling in variant interpretation and its potential to improve the accuracy and granularity of variant classifications.

Asset Subtitle

Presenting Author - Toby R. Manders, MD; Co-Author - Yuya Kobayashi, PhD; Co-Author - John Nicoludis, PhD; Co-Author - Arun Numpally, PhD; Co-Author - Flavia M. Facio, MS, CGC; Co-Author - Britt A. Johnson, PhD, FACMG; Co-Author - Keith Nykamp, PhD; Co-Author - Robert L. Nussbaum, MD; Co-Author - Alex Colavin, PhD;

Meta Tag

Clinical History

Genetic Diversity

Genetic Testing

Genomic Methodologies

Genotype-Phenotype Correlations

Natural History

Phenotype

Phenotypic delineation of disorders

Population Genetics

Risk Assessment

Co-Author Yuya Kobayashi, PhD

Co-Author John Nicoludis, PhD

Co-Author Arun Numpally, PhD

Co-Author Flavia M. Facio, MS, CGC

Co-Author Britt A. Johnson, PhD, FACMG

Co-Author Keith Nykamp, PhD

Co-Author Robert L. Nussbaum, MD

Co-Author Alex Colavin, PhD

Presenting Author Toby R. Manders, MD

Keywords

allele frequency model

variant interpretation

genetic variants

probabilistic graphical models

population allele frequency

Bayesian statistical frameworks

pathogenicity prediction

Genome Aggregation Database

research-eligible sample database

variant pathogenicity