FINSURF

FINSURF assigns functional impact scores to non-coding sequence variants in human regulatory regions using random forests to prioritize variants for clinical diagnostics and research interpretation.

Key Features:

• Variant Analysis: Evaluates Single Nucleotide Variants (SNVs), insertions, and deletions, classifying SNVs into transitions and transversions and scoring insertions and deletions based on their flanking or deleted bases.
• Functional Impact Prediction: Predicts the functional impact of non-coding variants using a random forest machine-learning approach trained with optimized selection of control variants to enhance predictive accuracy.
• Annotation Contribution Breakdown: Quantifies how genomic, functional, and evolutionary annotations contribute to each variant's overall score.
• Disease Gene Integration: Integrates lists of known or suspected disease genes to focus scoring on variants overlapping cis-regulatory elements linked to those genes.
• Graphical Representation: Produces graphical representations that illustrate the relative contributions of different annotations to a variant's score.

Scientific Applications:

• Disease Diagnosis and Research: Aids identification and prioritization of disease-causing non-coding mutations from whole-genome sequencing (WGS) that may be overlooked by coding-focused analyses.
• Personalized Medicine: Improves interpretation of WGS in personalized medicine by prioritizing candidate non-coding variants when protein-coding mutations are absent, thereby enhancing diagnostic yield.

Methodology:

Employs a random forest machine-learning framework that integrates diverse genomic data and trains on optimized control variant sets; validated across 30 diseases with known causative non-coding mutations, demonstrating high accuracy among top-ranked hits.