xQTLImp

xQTLImp imputes missing xQTL summary statistics (including eQTL, mQTL, and haQTL) from Z statistics by modeling variant association patterns to enhance discovery when individual-level genotypes and molecular traits are unavailable.

Key Features:

• Imputation Without Individual-Level Data: Operates on summary statistics such as Z statistics and does not require individual-level genotypes or molecular traits.
• Multivariate Gaussian Approximation: Models association statistics of variants linked to the same molecular trait using a multivariate Gaussian model and leverages linkage disequilibrium (LD) among variants for inference.
• Enhancement of Discovery Power: Imputes missing xQTL associations to reduce the lower bound on minor allele frequency (MAF) required for detection, enabling identification of novel xQTL signals.
• Validation Through Real Datasets: Validated on multiple real datasets, including a single-cell eQTL dataset, demonstrating high-accuracy rediscovery of significant eQTL associations.

Scientific Applications:

• Cross-omics QTL analysis: Imputation across eQTL, mQTL, and haQTL to extend discovery in genomics and multiomic studies.
• Studies lacking individual-level data: Enables xQTL discovery in cohorts or meta-analyses where only summary statistics are available.
• Low-frequency variant discovery: Facilitates detection of associations involving lower MAF variants by imputing missing statistics.

Methodology:

The software implements a C++ multivariate Gaussian approximation of variant association statistics that leverages linkage disequilibrium (LD) patterns among variants to impute missing Z statistics.