Refining fine-mapping: Effect sizes and regional heritability

Area of Science:

• Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Complex traits and diseases are influenced by numerous genetic variants, collectively explaining more phenotypic variance than individually associated variants.
• Increasing sample sizes enhance the detection and prioritization of specific variants linked to genomic regions and phenotypes.

Purpose of the Study:

• To extend the FINEMAP software for estimating effect sizes and regional heritability using a probabilistic model with few causal variants per region.
• To compare FINEMAP's performance against existing variance component (BOLT) and fixed-effect (HESS) models.

Main Methods:

• Simulated genomic regions using UK Biobank (UKB) data to assess precision and heritability decomposition.
• Applied FINEMAP, BOLT, and HESS to UKB plasma protein data (2,940 proteins) to evaluate regional heritability capture.

Main Results:

• FINEMAP demonstrated higher precision and more detailed regional heritability decomposition than BOLT and HESS, especially with few causal variants.
• FINEMAP identified an average of 2.5 causal variants per association signal, capturing 36% more regional heritability than the top associated variant.
• FINEMAP captured 13% and 40% more heritability than BOLT and HESS, respectively, in regions with significant heritability contributions.

Conclusions:

• FINEMAP provides a more refined approach to understanding the genetic architecture of complex traits by accounting for multiple causal variants.
• The extended FINEMAP offers improved estimation of regional heritability, outperforming established methods like BOLT and HESS in key scenarios.
• This work clarifies the relationships between FINEMAP, BOLT, and HESS, particularly for inferring variant-level genetic architecture.