RankVar: machine learning-based variant ranking and reinterpretation for rare genetic diseases

Area of Science:

• Genomics
• Bioinformatics
• Machine Learning

Background:

• Integrating biological knowledge and phenotype data with sequencing is crucial for identifying disease genes.
• Prioritizing causative variants from large datasets remains a significant challenge in genetic studies.

Purpose of the Study:

• To develop and evaluate RankVar, a machine learning algorithm for prioritizing causative variants in rare diseases.
• To leverage clinical notes and genome/exome sequencing data for improved variant prioritization.

Main Methods:

• RankVar employs a random forest classifier trained on approximately 1 million variants.
• The algorithm was tested on independent datasets including Mendelian diseases, birth defects, and autism spectrum disorders.
• Sequencing data and phenotype information were utilized for training and validation.

Main Results:

• RankVar achieved high top-10 variant accuracy across multiple datasets (e.g., 90.0% for CHOP).
• The algorithm successfully identified X-linked and Y-linked disease-causal variants.
• RankVar facilitated the reinterpretation of unsolved cases, identifying 61 candidate causal variants for hearing loss.

Conclusions:

• RankVar demonstrates superior performance compared to existing methods for variant prioritization.
• The algorithm accommodates diverse genetic models and X/Y chromosome variants.
• RankVar offers a valuable framework for genetic diagnosis, reinterpretation, and novel disease gene discovery.