POKEMON

POKEMON detects associations between rare missense variants and phenotypes by evaluating their three-dimensional spatial distribution within protein structures to increase power in rare-variant association testing.

Key Features:

• Structure-based evaluation: Evaluates rare missense variants based on their three-dimensional spatial distribution within protein structures rather than allele frequency.
• Protein-optimized kernel evaluation: Aggregates variant effects using a kernel-based approach (Protein Optimized Kernel Evaluation of Missense Nucleotides) in protein spatial context.
• Power for rare variants: Enhances statistical power to detect associations involving rare variants, including singletons, in whole-exome sequencing datasets such as ADSP WES.
• Spatial clustering analysis: Identifies localized clusters of rare variants within functional protein domains, for example ligand-binding domains.
• Replication support: Enables validation of spatial association signals in independent replication datasets.

Scientific Applications:

• Rare variant association testing in WES: Detects associations between rare missense variants and phenotypes in whole-exome sequencing datasets such as the Alzheimer’s Disease Sequencing Project (ADSP WES).
• Alzheimer’s disease gene discovery: Identified two known AD genes (TREM2, SORL1) and two novel candidate genes (DUSP18, CSF1R) associated with Alzheimer’s disease in ADSP WES analyses.
• Functional inference via spatial clustering: Maps spatial clusters of rare variants to protein domains to infer functional impacts, exemplified by a case-enriched DUSP18 cluster around its ligand-binding domain.

Methodology:

POKEMON implements a protein-optimized kernel evaluation that assesses spatial distributions of rare missense variants on protein structures and performs spatial clustering analysis on whole-exome sequencing data for association testing and replication.