PLATO

PLATO implements a filter-based framework to analyze genome-wide association study (GWAS) datasets and facilitate detection of gene-gene interactions and modest-effect genetic variants.

Key Features:

• Filter-Based System: PLATO consolidates multiple analytical techniques into a filter-based methodology to refine large genomic datasets for interaction analysis.
• Single Locus Filters: PLATO implements and evaluates single-locus filters as an initial step to select relevant genetic variants for downstream interaction testing.
• Analytical Filter Optimization: PLATO assessed 24 analytical filters for redundancy using a kappa score and grouped them into four distinct filter classes.
• MAX Statistic Implementation: PLATO computes a MAX statistic by taking the smallest p-value across the four selected filters for each single nucleotide polymorphism (SNP).
• Permutation Testing: PLATO uses permutation testing to empirically estimate p-values and assess statistical significance.
• Simulation Studies: PLATO evaluated statistical power and false positive rates through simulation studies comparing the combined filter approach to individual filters.

Scientific Applications:

• Detection of gene-gene interactions in GWAS: Prioritizes SNPs and variant sets for downstream epistasis and interaction analyses in complex trait and disease studies.
• Epistasis analysis of modest-effect variants: Enhances identification of genetic effects with modest impact sizes by aggregating multiple filter results via the MAX statistic.
• Assessment of filter performance: Uses simulation studies and permutation-derived empirical p-values to evaluate power and false positive rates of analytical filters.

Methodology:

Consolidation of multiple analytical filters into a filter-based system; implementation and evaluation of single-locus filters; assessment of 24 analytical filters with kappa score to identify redundancy and grouping into four filter classes; calculation of a MAX statistic by selecting the smallest p-value across the four filters per SNP; permutation testing to obtain empirical p-values; simulation studies to evaluate power and false positive rate.