AIVAR

AIVAR classifies genetic variants using a neural network to predict variant pathogenicity and support interpretation of sequence and functional assay data.

Key Features:

• Neural network architecture: Employs a neural network framework to process complex sequence-based features from high-throughput sequencing data for variant classification.
• Comparison with human experts: Validated against human expert classifications across multiple datasets, demonstrating high concordance.
• Integration with in silico tools: Can be used alongside computational predictors such as CADD (Combined Annotation Dependent Depletion) and PhyloP to augment variant interpretation.
• Functional assay correlation: Evaluated against functional assay data including saturation genome editing (SGE), with reported analyses showing non-significant concordance with SGE function scores.

Scientific Applications:

• Variant classification: Assigns likely pathogenicity to variants, including variants of uncertain significance, to aid interpretation.
• Clinical genomics and diagnostics: Supports identification of pathogenic mutations for genomic research and clinical diagnostic workflows.
• Functional genomics integration: Bridges high-throughput sequencing data and functional assays to enable multi-dimensional genomic analyses.

Methodology:

A neural network model is trained on large datasets of known genetic variants using sequence-based features and, where available, functional assay data to learn patterns associated with variant pathogenicity.