Polygenic modelling of treatment effect heterogeneity

Area of Science:

• Genetics
• Epidemiology
• Pharmacology

Background:

• Mendelian randomization (MR) uses genetic variants to infer causal effects of risk factors on outcomes.
• Pharmacological treatments aim to modify risk factors, but treatment effects can vary across individuals.
• Identifying subgroups with differential treatment response is crucial for personalized medicine.

Purpose of the Study:

• To develop and illustrate a method for detecting gene-gene interactions to identify genetic subgroups with differential treatment response.
• To apply this method to investigate effect heterogeneity in statin treatment on low-density lipoprotein cholesterol.

Main Methods:

• Utilizes a "pharmacogenetic" variant as a proxy for pharmacological treatment.
• Tests for interactions between the pharmacogenetic variant and other genetic variants (effect modifiers) associated with a risk factor.
• Employs a random forest of interaction trees to construct a polygenic response score when individual effect modifiers are not found.

Main Results:

• The method can identify genetic subgroups with varying predicted treatment responses.
• Demonstrates application in identifying heterogeneity of statin effects on LDL cholesterol, suggesting potential for personalized statin therapy.

Conclusions:

• This approach enhances Mendelian randomization by incorporating gene-gene interactions to predict treatment efficacy in specific genetic subgroups.
• The polygenic response score offers a novel tool for personalized treatment response prediction, analogous to polygenic risk scores.