IDRMutPred

IDRMutPred predicts disease-associated nonsynonymous single nucleotide variants (nsSNVs) within intrinsically disordered regions (IDRs) of proteins to prioritize variants relevant to human disease.

Key Features:

• Target region: Focuses specifically on nsSNVs located in intrinsically disordered regions (IDRs) of proteins.
• Machine learning approach: Employs a machine learning-based approach tailored for IDR nsSNVs.
• Feature set: Uses 17 optimal features derived from sequence alignments, protein annotations, hydrophobicity indices, and disorder scores.
• Algorithms: Implements three ensemble learning algorithms trained exclusively on datasets comprising IDR nsSNVs.
• Performance: Achieves area under the curve (AUC) between 0.856 and 0.868 and Matthews correlation coefficient (MCC) between 0.713 and 0.737 across two testing datasets and outperforms 17 general-purpose predictors.

Scientific Applications:

• Variant prioritization in disease studies: Prioritizes candidate disease-associated nsSNVs within IDRs for studies of the genetic basis of human disease.
• Benchmarking IDR-specific predictors: Enables comparative evaluation of variant effect predictors specifically in intrinsically disordered regions.

Methodology:

Applies three ensemble learning algorithms trained on IDR nsSNV datasets using 17 features derived from sequence alignments, protein annotations, hydrophobicity indices, and disorder scores, with performance assessed by AUC and MCC on two testing datasets.