Basic/Translational Science -> Genomics: Translational D-AB02 - Machine Learning and Computational Modeling: A Step Towards Precision Medicine? (ID 16) Abstract Plus

D-AB02-02 - Use Of Machine Learning To Identify High Risk Variants Of Uncertain Significance In Lamin A/c Cardiomyopathy (ID 723)

Disclosure
 J. Bennett: Nothing relevant to disclose.
Audio File Upload

Abstract

Background: Mutations in Lamin A/C result in a spectrum of clinical disease, including cardiomyopathy and conduction system disease. Variants implicated in cardiomyopathy are often found in the rod domain of the gene. Stratification of clinical risk and classification of LMNA variants is difficult, with little benign variation and a high rate of variants of uncertain significance (VUS).
Objective: To reclassify a familial variant as pathogenic and use machine learning to identify VUS in the LMNA rod domain with increased likelihood of pathogenicity.
Methods: Missense variants in the LMNA gene rod domain were identified from ClinVar and Gnomad, and in silico predictions of conservation and pathogenicity were recorded. Uniform Manifold Approximation and Projection was used to project in silico predictions of pathogenicity and conservation for single nucleotide missense variants in the LMNA rod domain. Clustering was performed using Density-Based Spatial Clustering of Applications with Noise (DBSCAN). Genetic variants were evaluated using American College of Medical Genetics and Genomics guidelines.
Results: We discovered a novel missense variant (D136E) in the LMNA rod domain segregating with cardiac disease in a multi-generation pedigree and accordingly reclassified it as likely pathogenic. Using machine learning on in silico predictions of pathogenicity, two clusters of variation in the rod domain were identified. Known pathogenic variants significantly more frequent in Cluster 2 than Cluster 1 (3.1% of total variants in Cluster 1 vs. 28.6% in Cluster 2, p=1.68e-7). Variants were all rare, with population allele frequency <0.005. Of 186 VUS in Cluster 2, 115 (61.8%) were absent in control population databases. Given the high constraint with lack of benign variation within that region of the gene, they would be classified as likely pathogenic if segregating with disease within a family.
Conclusion: Computational and clinical data were used to reclassify a LMNA genetic variant as disease-causing without the need for additional testing of family members. Machine learning identified a variant cluster at high risk to be disease-causing; familial testing for patients with high risk VUS may assist in classification of variant pathogenicity and disease risk.
Collapse