GWAS with longitudinal phenotypes: performance of approximate procedures.

Area of Science:

• Genetics
• Biostatistics
• Computational Biology

Background:

• Standard linear mixed model (LMM) fitting for genome-wide association studies (GWAS) with longitudinal data is computationally intensive.
• Accelerating these computations is crucial for efficient genetic analysis.

Purpose of the Study:

• To further investigate the properties of the conditional two-step (CTS) approach for analyzing GWAS with longitudinal data.
• To compare the performance of CTS with a related two-step procedure.

Main Methods:

• The conditional two-step (CTS) approach involves fitting a reduced LMM without SNP terms, followed by regressing estimated random slopes on SNPs.
• The study employed analytical derivations and simulations to evaluate CTS performance.
• Comparison was made against a standard two-step procedure.

Main Results:

• Analytically shown that for balanced data, CTS P-values are equivalent to LMM P-values under mild assumptions.
• Simulations demonstrate that CTS does not inflate the type I error rate for unbalanced data.
• CTS results in only a minimal loss of statistical power compared to LMM.

Conclusions:

• The conditional two-step (CTS) approach provides a computationally efficient and accurate method for GWAS with longitudinal data.
• CTS is a viable alternative to standard LMM fitting, especially for large datasets or limited computational resources.
• The method maintains statistical rigor while drastically reducing analysis time.