Detecting gene-environment interactions to guide personalized intervention: Boosting distributional regression for...

Area of Science:

• Genetics
• Biostatistics
• Computational Biology

Background:

• Current polygenic risk scores (PRS) primarily model trait means, overlooking genetic contributions to phenotypic variance.
• Understanding genetic influence on variance is crucial for gene-environment interaction (GxE) studies.
• Existing methods lack efficient approaches for high-dimensional genetic data.

Purpose of the Study:

• To introduce snpboostlss, a novel algorithm for jointly modeling the mean and variance of quantitative phenotypes using sparse polygenic models.
• To enhance computational efficiency for large-scale genotype datasets.
• To investigate GxE by analyzing phenotypic variance.

Main Methods:

• Developed snpboostlss, a cyclical gradient boosting algorithm for Gaussian location-scale models.
• Implemented a batch-wise approach focusing on relevant variants for computational efficiency.
• Applied the algorithm to UK Biobank data, analyzing statin therapy effects on low-density lipoprotein (LDL) and body mass index (BMI) interactions with lifestyle factors.

Main Results:

• Identified significant interactions between statin usage and PRS for phenotypic variance in both cross-sectional and longitudinal analyses.
• Observed a more substantial statin treatment effect in individuals with higher PRS for phenotypic variance, indicating GxE.
• Demonstrated significant interactions between PRS for variance and physical activity/sedentary behavior for BMI.

Conclusions:

• The proposed snpboostlss algorithm effectively derives PRS for phenotypic variance, capturing GxE.
• These variance-based PRS can identify individuals likely to benefit from targeted environmental interventions.
• This approach holds potential for personalized medicine and lifestyle recommendations.