IDminer

IDminer analyzes a curated cohort of intellectual disability patients to identify pathogenic genes, pathogenic variant sites, and genotype–phenotype associations that elucidate molecular perturbations underlying intellectual disability.

Key Features:

• Curated cohort: Systematically curated phenotyping cohort comprising 3,803 patients with intellectual disability.
• Variant and gene catalog: Identification of 704 pathogenic genes, 3,848 pathogenic sites, and 2,075 standard phenotypes.
• Phenotypic heterogeneity analysis: Quantifies clinical heterogeneity and reports a positive correlation between number of phenotypes and number of affected patients.
• Mutation-type and gene distribution analysis: Reports mutation type biases and that the top 44 genes account for nearly 40% of cases in the cohort.
• Phenotype co-occurrence networks: Enriched co-occurrent phenotype networks associated with each gene to reveal phenotype convergence and support gene prioritization.
• IDpred machine learning predictor: A machine learning–based predictor (IDpred) for pathogenic genes with reported AUC = 0.978 using 10-fold cross-validation.
• Comprehensive database: Hosts a comprehensive database of the ID phenotyped cohort and curated ID data.

Scientific Applications:

• Pathogenic gene prioritization and prediction: Prioritizes and predicts ID-associated pathogenic genes using IDpred and curated cohort data.
• Genotype–phenotype correlation: Investigates relationships between genetic variants and standardized phenotypes to elucidate molecular perturbations in intellectual disability.
• Phenotype convergence and mechanism inference: Uses co-occurrence networks to identify convergent phenotypes linked to specific genes for insights into cognitive developmental disorder mechanisms.
• Resource for diagnosis and research: Provides a curated dataset and variant/gene catalog to support diagnostic interpretation and research into the pathogenesis of intellectual disability.

Methodology:

Systematic curation of a 3,803-patient phenotyping cohort; identification of 704 pathogenic genes, 3,848 pathogenic sites, and 2,075 standard phenotypes; enriched co-occurrent phenotype network analysis per gene; machine learning predictor IDpred evaluated by 10-fold cross-validation (AUC = 0.978).