SuSiE 2.0: improved methods and implementations for genetic fine-mapping and phenotype prediction.

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Sum of Single Effects regression (SuSiE) is a popular method for genetic fine-mapping.
• The original SuSiE implementation has limitations in extensibility and performance.

Purpose of the Study:

• To present SuSiE 2.0, a redesigned and enhanced version of the SuSiE algorithm.
• To introduce SuSiE-ash, a novel method for improved calibration in genetic fine-mapping.
• To evaluate the performance and utility of SuSiE 2.0 and SuSiE-ash.

Main Methods:

• Modular redesign of the SuSiE algorithm for improved extensibility.
• Implementation of speed enhancements for summary statistics applications.
• Development and application of the SuSiE-ash method for handling complex genetic signals.
• Simulations and real data benchmarks across diverse genetic architectures.

Main Results:

• SuSiE 2.0 offers up to 5x speed improvements.
• SuSiE-ash demonstrates improved calibration, achieving 1.5-3x FDR reduction in complex polygenic settings while maintaining power.
• SuSiE-based methods show effectiveness for TWAS (Transcriptome-Wide Association Studies) prediction.

Conclusions:

• SuSiE 2.0 provides a more extensible and performant framework for genetic fine-mapping.
• SuSiE-ash significantly improves fine-mapping calibration under complex genetic architectures.
• SuSiE methods are valuable tools for genetic fine-mapping and TWAS prediction.