A quadratically regularized functional canonical correlation analysis for identifying the global structure of...

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Investigating pleiotropic effects of genetic variants enhances statistical power and disease understanding.
• Current methods for multiple phenotype association analysis lack breadth and depth.
• Extracting informative features from high-dimensional genotype and phenotype data is crucial for pleiotropic analysis.

Purpose of the Study:

• To develop a novel statistical method for high-dimensional pleiotropic analysis.
• To improve the extraction of correlation information and reduce data dimensions in genetic data.
• To overcome limitations in current statistical methods and computational algorithms for genetic pleiotropic analysis.

Main Methods:

• Proposed a quadratically regularized functional canonical correlation analysis (QRFCCA) method.
• Combined quadratically regularized matrix factorization, functional data analysis, and canonical correlation analysis (CCA).
• Validated the method using large-scale simulations and the TwinsUK whole genome sequencing dataset.

Main Results:

• QRFCCA demonstrated significantly higher statistical power compared to ten competing statistics in simulations.
• QRFCCA maintained appropriate type 1 error rates.
• Identified 79 genes with rare variants and 67 genes with common variants associated with 46 traits in the TwinsUK study.

Conclusions:

• QRFCCA substantially outperforms existing methods in identifying genetic variant-trait associations.
• The developed method advances statistical and computational approaches for genetic pleiotropic analysis.
• QRFCCA offers a powerful tool for understanding complex genetic disease structures and designing targeted treatments.