Evaluation of Bayesian Linear Regression Models as a Fine Mapping tool

Abstract Our aim was to evaluate Bayesian Linear Regression (BLR) models with BayesC and BayesR priors as a fine mapping tool and compare them to the state-of-the-art external models: FINEMAP, SuSIE-RSS, SuSIE-Inf and FINEMAP-Inf. Based on extensive simulations, we evaluated the different models based on F1classification score. The different models were applied on quantitative and binary UK Biobank (UKB) phenotypes and evaluated based upon predictive accuracy and features of credible sets (CSs). We used over 533K genotyped and 6.6 million imputed single nucleotide polymorphisms (SNPs) for simulations and UKB phenotypes respectively, from over 335K UKB White British Unrelated samples. We simulated phenotypes from low (GA1) to moderate (GA2) polygenicity, heritability (h2) of 10% and 30%, causal SNPs (π) of 0.1% and 1% sampled genome-wide, and disease prevalence (PV) of 5% and 15%. Single marker summary statistics and in-sample linkage disequilibrium were used to fit models in regions defined by lead SNPs. BayesR improved the F1score, averaged across all simulations, between 27.26% and 13.32% relative to the external models. Predictive accuracy quantified as variance explained (R2), averaged across all the UKB quantitative phenotypes, with BayesR was decreased by 5.32% (SuSIE-Inf) and 3.71% (FINEMAP-Inf), and was increased by 7.93% (SuSIE-RSS) and 8.3% (BayesC). Area under the receiver operating characteristic curve averaged across all the UKB binary phenotypes, with BayesR was increased between 0.40% and 0.05% relative to the external models. SuSIE-RSS and BayesR, demonstrated the highest number of CSs, with BayesC and BayesR exhibiting the smallest average median size CSs in the UKB phenotypes. The BLR models performed similar to the external models. Specifically, BayesR’s performance closely aligned with SuSIE-Inf and FINEMAP-Inf models. Collectively, our findings from both simulations and application of the models in the UKB phenotypes support that the BLR models are efficient fine mapping tools..

Medienart:	Preprint

Erscheinungsjahr:	2024
Erschienen:	2024

Enthalten in:	bioRxiv.org - (2024) vom: 16. Apr. Zur Gesamtaufnahme - year:2024

Sprache:	Englisch

Beteiligte Personen:	Shrestha, Merina [VerfasserIn] Bai, Zhonghao [VerfasserIn] Gholipourshahraki, Tahereh [VerfasserIn] Hjelholt, Astrid J. [VerfasserIn] Hu, Sile [VerfasserIn] Kjølby, Mads [VerfasserIn] Rohde, Palle D. [VerfasserIn] Sørensen, Peter [VerfasserIn]

Links:	Volltext [kostenfrei]

Themen:	570 Biology

doi:	10.1101/2023.09.01.555889

funding:
Förderinstitution / Projekttitel:

PPN (Katalog-ID):	XBI040745384