AVADA

AVADA automates retrieval of variant-level pathogenic evidence from full-text primary literature to accelerate cataloging of disease-causing genetic variants for monogenic disease research and clinical diagnosis.

Key Features:

• Machine learning and NLP: AVADA uses machine learning and natural language processing to scan and interpret full-text primary literature for variant evidence.
• Variant-to-coordinate mapping: It extracts mentions of pathogenic genetic variants and converts them into precise genomic coordinates.
• High retrieval efficiency: AVADA automatically retrieves approximately 60% of likely disease-causing variants listed in the Human Gene Mutation Database (HGMD), representing a 4.4-fold improvement over existing open-source automated variant extractors.
• Comprehensive variant coverage: The system includes over 60,000 likely disease-causing variants present in HGMD but absent from ClinVar.
• Diagnostic utility: In a cohort of 245 diagnosed patients, AVADA provided evidence for 18 additional diagnostic variants beyond those identified by ClinVar, compared with two additional variants found using the best current automated methods.
• Full-text literature focus: AVADA operates on full-text primary literature and accelerates retrieval relative to manual searches through resources such as PubMed or Google Scholar.

Scientific Applications:

• Pathogenic variant cataloging: Automated extraction of literature-based evidence to populate and expand variant databases such as HGMD and ClinVar.
• Clinical diagnosis support: Provision of variant-level evidence to aid interpretation and diagnosis of monogenic disease patients.
• Database curation: Identification of variants absent from ClinVar and provision of supporting literature citations for curation efforts.
• Monogenic disease research: Facilitation of discovery and characterization of disease-causing variants and genes from the primary literature.

Methodology:

AVADA applies machine learning–based natural language processing to systematically scan and interpret full-text primary literature, extract mentions of pathogenic genetic variants, and map those mentions to genomic coordinates.