Performance Comparison of LASSO Variants with Genome-Wide Association Studies (GWAS)

GWAS are popular approaches to associate genetic variations on a population of individuals with particular traits. Despite widely applications, GWAS are heavily depend on the accuracy of statistical models and they have been criticized for missing important genetic markers. Least absolute shrinkage and selection operator (LASSO) and its variants have become an important machine learning methods to find significant features from complex datasets. The regression models used in LASSO make it so useful for selecting features from the dataset where there are many features but only a few relevant ones. In this study, we investigate how LASSO-based methods can be used as an alternative of GWAS to find significant genetic markers. This study provides guidance on selecting LASSO-based programs with different data used in selecting significant SNPs.