GWiS

GWiS identifies independent genetic variant effects within genes using greedy Bayesian model selection and permutation-based p-value correction to produce gene-level significance metrics for genome-wide association studies.

Key Features:

• Greedy Bayesian Model Selection: Uses greedy Bayesian model selection to identify independent genetic effects within a gene and integrate multiple variant contributions into a single statistical signal.
• Permutation Tests for P-value Correction: Employs permutation tests to generate p-values that account for the number of independent tests both genome-wide and within each genetic locus.
• Enhanced Detection of Validated Associations: Demonstrated improved detection of validated associations in large GWAS datasets; for example, in a study of 2.5 million SNPs and up to 8,000 individuals measured for electrocardiography (ECG) parameters, GWiS identified more significant associations than conventional methods.
• Systematic Assessment of Genetic Effects: Quantifies the number of independent effects within genes and reports the fraction of loci with multiple independent effects (observed as 35%–50% in the referenced study).
• Generalizability and Power Retention: Generalizable to study designs beyond standard GWAS and retains analytical power for low-frequency alleles.
• Compatibility with Pathway-Based Meta-Analysis: Produces gene-based p-values that are directly compatible with pathway-based meta-analysis.

Scientific Applications:

• Complex trait locus dissection: Dissects genes harboring multiple causal variants to improve understanding of the genetic architecture underlying complex traits.
• Large-scale GWAS analysis: Enhances identification of validated associations in large SNP datasets (millions of SNPs across thousands of individuals), including studies of electrocardiography (ECG) parameters.
• Downstream integration: Provides gene-level statistics to support pathway-based meta-analysis, prioritization of potential therapeutic targets, and refinement of genetic risk prediction.

Methodology:

Applies greedy Bayesian model selection to identify independent variant effects within genes, uses permutation tests to generate p-values correcting for genome-wide and per-locus multiple testing, and aggregates identified independent effects into gene-level statistics while quantifying the number of independent effects per gene.