ASCARIS

ASCARIS constructs numerical representations of single amino acid variations (SAVs) to support prediction and analysis of their functional effects in proteins.

Key Features:

• Positional Feature Annotation: Represents SAVs using 30 different types of positional annotations, including active sites, lipidation and glycosylation sites, calcium/metal/DNA binding regions, and inter/transmembrane segments.
• Protein Structure-Based Representation: Integrates structural features alongside sequence annotations to capture direct and spatial correspondences between variations and protein structure.
• Physicochemical Properties Integration: Incorporates amino acid physicochemical properties into SAV representations to reflect biochemical implications of substitutions.
• Statistical Analysis and Feature Correlation: Provides statistical analyses to assess relationships between annotated features and variation consequences and to quantify each feature's contribution to functional impact prediction.
• Variant Effect Prediction Models: Uses the generated SAV representations as input for training variant effect prediction models.
• Ablation and Comparative Studies: Includes ablation studies and comparisons with state-of-the-art methods to evaluate feature importance and relative performance.

Scientific Applications:

• Genetic disease research: Identifies and characterizes deleterious SAVs relevant to genetic diseases.
• Personalized medicine: Supports interpretation of patient-specific SAVs for clinical and therapeutic decision-making.
• Drug discovery: Assesses SAV impacts on protein function to inform target validation and drug design.
• Multi-omics integration: Serves as a representation layer for building comprehensive models that integrate multi-omics data.

Methodology:

Constructs reusable numerical representations of SAVs by combining sequence- and structure-based annotations with amino acid physicochemical properties, applies statistical analyses to relate features to variation consequences, and uses the representations to train and evaluate variant effect prediction models including ablation and comparative studies.