DISSCO

DISSCO performs imputation of association summary statistics while accounting for covariates to improve accuracy of genetic association analyses when individual-level genotype data are unavailable.

Key Features:

• Direct imputation of summary statistics: Enables imputation of association summary statistics when individual-level genotype data are not available.
• Covariate adjustment: Accounts for confounding covariates rather than assuming equivalence between genotype correlations and summary statistic correlations.
• Partial correlation-based inference: Utilizes partial correlation among genotypes controlling for covariates to perform imputation.
• Comparison to existing methods: Explicitly addresses limitations of DIST and ImpG-Summary/LD that assume correlations among summary statistics mirror genotype correlations.
• Implementation: Implemented as a JAVA application.
• Performance on low-frequency variants: Demonstrates improved accuracy for variants with minor allele frequency <5%, reducing absolute deviations from true values by 3.9–15.2% versus correlation-based methods.

Scientific Applications:

• Association studies in admixed populations: Imputes summary statistics while adjusting for population-specific covariates in admixed cohorts.
• Studies with confounding covariates: Enables more accurate imputation in analyses where environmental or other covariates affect genotype–phenotype associations.
• Meta-analyses using summary statistics: Supports meta-analytic and association studies that rely on summary-level data rather than individual-level genotypes.

Methodology:

Imputes summary statistics using partial correlation among genotypes while controlling for covariates; implemented in JAVA and evaluated against DIST and ImpG-Summary/LD with reported reductions in absolute deviations for MAF <5%.