Univariate-guided sparse regression for Biobank-scale high-dimensional omics data.

UniLasso achieves sparse, interpretable polygenic risk scores with performance comparable to standard Lasso in UK Biobank data.

• Why it matters: High-dimensional genomic data require efficient methods to identify relevant variants and improve model interpretability without sacrificing predictive accuracy.
• What they did: The authors developed uniLasso, a two-stage penalized regression leveraging univariate coefficients, and extended it to incorporate external summary statistics, applying it to over one million variants in the UK Biobank.
• The result: UniLasso produces sparser models with fewer variants, maintains predictive performance similar to Lasso, and remains competitive with methods like PRS-CS and lassosum2, enabling more interpretable genetic risk prediction.