kTWAS

kTWAS integrates transcriptome-wide association study (TWAS)-style feature selection with Sequence Kernel Association Test (SKAT)-like kernel-based testing to aggregate contributions of multiple genetic variants and improve detection of genotype-phenotype associations.

Key Features:

• Elastic Net-based feature selection (TWAS-like): Selects genetic variants based on their impact on gene expression and produces pretrained linear combinations of variants for association mapping.
• SKAT-like kernel-based score testing: Applies kernel functions to model genetic similarity and to model genotypic and phenotypic variance, enabling detection of non-linear genotype-phenotype effects.
• Integration of complementary approaches: Combines TWAS-like feature pruning with SKAT-like kernel modeling to aggregate variant contributions within focal regions.

Scientific Applications:

• Improved statistical power: Demonstrates increased power to detect associations compared with traditional TWAS and multiple SKAT-based protocols in simulations.
• Disease gene discovery in cohort data: Applied to WTCCC genotyping array data and MSSNG (Autism) sequence data to identify disease-associated genes.

Methodology:

Applies Elastic Net-based feature selection akin to TWAS to generate pretrained linear combinations of variants, followed by a SKAT-like kernel-based score test using kernel functions to model genetic similarity and capture non-linear genotype-phenotype relationships while modeling genotypic and phenotypic variance.