Enhanced genetic fine mapping accuracy with Bayesian Linear Regression models in diverse genetic architectures
Merina Shrestha, Zhonghao Bai, Tahereh Gholipourshahraki, Astrid J. Hjelholt, Sile Hu,Mads Kjolby, Palle Duun Rohde, Peter Sørensen
Abstract
We evaluated Bayesian Linear Regression (BLR) models with BayesC and BayesR priors as statistical genetic fine-mapping tools, comparing their performance to established methods such as FINEMAP and SuSiE. Through extensive simulations and analyses of UK Biobank (UKB) phenotypes, we assessed F1 classification scores and predictive accuracy across models.
Introduction
To better understand the genetic architecture of complex traits and multifactorial diseases, it is essential to identify the genetic variants that are most likely causal or in linkage disequilibrium (LD) with the causal variants. Genome-wide association studies (GWAS) often identify numerous associated variants due to long-range LD, which complicates statistical inference [1].
Materials and method
UKB genetic data
UKB genotyped genetic variants were used for the simulation study whereas imputed genetic variants were used in analysis of UKB phenotypes. To obtain a genetic homogeneous study population we restricted our analyses to unrelated British Caucasians and excluded individuals with more than 5,000 missing variants or individuals with autosomal aneuploidy. Remaining (n = 335,532) White British unrelated individuals (WBU) were used for analyses.
Results
Performance in simulation scenarios
Comparison between methods.
We evaluated the mean performance (± standard error) of several fine-mapping methods across multiple metrics. Statistical significance was assessed using one-sample t-tests comparing each method’s mean performance to the overall metric average, as well as ANOVA on a linear model in which the performance metric (e.g., F1 score) was regressed on method, while adjusting for other simulation design parameters (Figs 4 and S3–S10 Figs).
Discussion
Here we aimed to assess the efficiency of BLR models with BayesC and BayesR priors as a fine mapping tool. We applied these models in simulations, and on empirical data from UKB using GWAS summary statistics.
Acknowledgments
This research has been conducted using the UK Biobank Resource under application number 96479.
Citation: Shrestha M, Bai Z, Gholipourshahraki T, Hjelholt AJ, Hu S, Kjolby M, et al. (2025) Enhanced genetic fine mapping accuracy with Bayesian Linear Regression models in diverse genetic architectures. PLoS Genet 21(7): e1011783.
https://doi.org/10.1371/journal.pgen.1011783
Editor: Gao Wang, Columbia Presbyterian Medical Center: Columbia University Irving Medical Center, UNITED STATES OF AMERICA
Received: April 5, 2024; Accepted: June 23, 2025; Published: July 30, 2025
Copyright: © 2025 Shrestha et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: S2–S5 Tables contain the processed fine-mapping results for all methods across all simulations and can be used to reproduce the figures and results presented in this revised manuscript. Example scripts for the fine-mapping procedures involving the evaluated BLR models are available at: https://psoerensen.github.io/gact/Document/Finemapping_bayesian_linear_regression_simulated_data.html. Functions for simulating the different scenarios used in this study, performing fine-mapping, and generating credible sets are implemented in the R package qgg, which is available at https://psoerensen.github.io/qgg/. The genotype and phenotype data used in our analyses are available from UK Biobank (https://www.ukbiobank.ac.uk/).
Funding: MK, PDR, and PS obtained funding from Novo Nordisk Foundation through the drug discovery platform, Open Discovery Innovation Network (ODIN) under grant number “NNF20SA0061466”. This funding aims to foster collaboration between universities and companies promoting long-term benefits of innovation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
