Potential application of elastic nets for shared polygenicity detection with adapted threshold selection.

Area of Science:

• Genetics
• Bioinformatics
• Statistical Genomics

Background:

• Many disorders have a polygenic basis, involving numerous single nucleotide polymorphisms (SNPs) with small effects.
• Polygenic overlap, where risk alleles are shared across genetic loci, is observed between various disorders.
• Current methods for detecting polygenic overlap, like rank-ordering p-values from genome-wide association studies (GWASs), have limitations.

Purpose of the Study:

• To explore the utility of elastic nets, a high-dimensional variable selection technique, for detecting shared polygenicity.
• To adapt elastic nets for use with polygenic scores as the dependent variable in GWAS analysis.
• To compare the performance of elastic net approaches against traditional rank-ordering methods for identifying shared polygenicity.

Main Methods:

• Developed and adapted elastic net models with polygenic scores as the dependent variable.
• Investigated appropriate methods for selecting the penalty parameter in elastic net models for GWAS data.
• Utilized synthetic datasets for theoretical understanding and extensive simulations to compare elastic nets with rank-ordering approaches.

Main Results:

• Elastic net approaches were theoretically grounded and illustrated using synthetic data.
• Simulation studies demonstrated that one elastic net approach outperforms rank-ordering when SNP correlations are high.
• The methods were applied to real datasets, highlighting their capabilities and limitations in detecting shared polygenicity.

Conclusions:

• Elastic nets offer a viable and potentially superior strategy for detecting shared polygenicity, particularly in complex genetic scenarios.
• The adapted elastic net methods provide a novel tool for dissecting the genetic architecture of overlapping disorders.
• Further application of these methods can enhance our understanding of shared genetic risk factors across diseases.