ReMM score

ReMM score predicts regulatory potential of genomic positions to prioritize non-coding variants from whole-genome sequencing data in Mendelian disease studies.

Key Features:

• Regulatory Variant Scoring: Scores genomic positions for regulatory probability and distinguishes deleterious from neutral variants using a machine learning framework tailored for highly imbalanced datasets.
• Imbalanced-data Algorithm (hyperSMURF): Employs the hyperSMURF algorithm specifically designed to handle highly imbalanced variant classification problems.
• Integration with Genomiser Framework: Operates within the Genomiser framework and integrates existing methods such as CADD alongside bespoke machine-learning approaches.
• Data Integration: Leverages curated regulatory variants, proxy-neutral variants, allele frequency, regulatory sequences, and chromosomal topological domains as input features.
• Phenotypic Relevance: Associates scored regulatory variants with Mendelian disease relevance to aid prioritization of phenotypically relevant variants.

Scientific Applications:

• Identifying Causal Regulatory Variants: Ranks causal regulatory variants as top candidates in a reported 77% of simulated whole genomes.
• Mendelian Disease Research: Facilitates discovery and interpretation of non-coding variants associated with Mendelian disorders from whole-genome sequencing data.

Methodology:

Uses hyperSMURF, a machine-learning algorithm for highly imbalanced datasets, within a Genomiser-integrated framework combining CADD and bespoke machine-learning models and input features including curated regulatory variants, proxy-neutral variants, allele frequency, regulatory sequences, and chromosomal topological domains applied to whole-genome sequencing data.