Genome-wide Analysis of Large-scale Longitudinal Outcomes using Penalization -GALLOP algorithm.

Area of Science:

• Genetics
• Computational Biology
• Statistical Genetics

Background:

• Genome-wide association studies (GWAS) analyze genetic variations linked to traits.
• Longitudinal phenotypes are crucial for understanding trait changes over time.
• Mixed models are standard for analyzing correlated longitudinal data in GWAS, but computationally intensive.

Purpose of the Study:

• To develop a significantly faster algorithm for GWAS on longitudinal data.
• To address the computational challenges of fitting mixed models to millions of SNPs for thousands of individuals.
• To enable efficient identification of genetic variations affecting trait trajectories.

Main Methods:

• A novel algorithm solving an equivalent penalized least squares problem.
• Efficient variance computation and use of factorizations/transformations to avoid matrix inversion.
• Leveraging bordered systems to reuse precomputed components for mixed models.

Main Results:

• The algorithm achieves speed-ups of several orders of magnitude compared to existing methods.
• Analysis is approximately 1000 times faster than the R package lme4.
• The method provides highly similar coefficients and p-values to established approaches.

Conclusions:

• The new algorithm dramatically improves the computational efficiency of GWAS for longitudinal data.
• This advancement facilitates large-scale genetic analyses of trait changes over time.
• An R implementation is provided for practical application.