UMD-Predictor

UMD-Predictor predicts the pathogenicity of cDNA substitutions across all human transcripts by integrating biochemical, splicing, protein-domain, population frequency, and conservation data to prioritize variants from whole-exome sequencing (WES) and next-generation sequencing (NGS) analyses.

Key Features:

• Combinatorial integration: Integrates multiple data points to assess the potential pathogenicity of cDNA substitutions across all human transcripts.
• Biochemical properties: Evaluates biochemical characteristics of mutant and wild-type residues to predict functional impact.
• Splicing signal impact: Assesses how genetic changes may affect splicing signals.
• Protein domain localization: Analyzes variant localization within protein domains to infer functional consequences.
• Population variation frequency: Incorporates global population frequency data to distinguish common polymorphisms from rare mutations.
• Conservation analysis: Uses the BLOSUM62 substitution matrix and conservation across 100 species as evolutionary evidence of functional importance.
• Transcript-wide assessment: Applies predictions to all human transcripts for comprehensive variant evaluation.
• Performance benchmarking: Was compared with seven leading prediction tools using over 140,000 annotated variations and reported superior accuracy, specificity, Matthews correlation coefficient, and diagnostic odds ratio.
• Prioritization efficiency: Produces a shortened, prioritized list of candidate mutations to reduce downstream analysis burden.
• Pipeline integration: Provides web services for integration into bioinformatics pipelines used for NGS analyses.

Scientific Applications:

• Variant prioritization in WES/NGS: Prioritizes candidate pathogenic variants from whole-exome and other NGS datasets for downstream analysis.
• Gene discovery: Facilitates identification of disease-associated genes by highlighting likely pathogenic substitutions.
• Clinical diagnostics support: Assists clinical interpretation by distinguishing likely pathogenic mutations from benign polymorphisms.
• Conservation and functional inference: Provides conservation-based evidence to support functional annotation of variants.

Methodology:

Combinatorial integration of biochemical property comparison (mutant versus wild-type), splicing signal impact analysis, protein domain localization, population frequency incorporation, and conservation scoring using BLOSUM62 across 100 species applied to cDNA substitutions across all human transcripts, with performance evaluated against over 140,000 annotated variations and seven other prediction tools using accuracy, specificity, Matthews correlation coefficient, and diagnostic odds ratio.