Genome-wide linkage analysis of blood pressure under locus heterogeneity

Xinqun Yang, Kai Wang, Jian Huang, Veronica J. VielandBMC 2003 · DOI

TL;DR

This paper introduces a new statistical method to identify genetic regions (loci) influencing complex traits like blood pressure, specifically accounting for "locus heterogeneity" – the idea that different families might have different genetic causes for the same trait. They applied this method to the Framingham Heart Study data and found four promising genetic markers, two of which align with previous findings and two that appear novel.

What Problem Does This Paper Address?

High blood pressure (BP) is a major risk factor for cardiovascular disease and mortality. It's a "complex trait," meaning it's influenced by multiple genes and environmental factors. A key challenge in mapping the genes responsible for such traits is "locus heterogeneity." This means that within a population, different families might have different genetic variants at different locations in the genome that contribute to their high blood pressure. Traditional linkage analysis methods often assume a single genetic cause across all families. This assumption can reduce the power to detect true genetic signals, especially when heterogeneity is present. The authors aim to develop and validate a statistical approach that can overcome this limitation and increase the power to detect genetic loci for complex traits.

Methods

The core of this paper is a novel statistical model designed to map quantitative trait loci (QTLs) while explicitly allowing for locus heterogeneity. This model generalizes existing heterogeneity models for qualitative traits to continuous traits like blood pressure.

Here's a breakdown of their approach:

  • Generalized Heterogeneity Model: They propose a likelihood-based model that considers the probability of a quantitative trait locus (QTL) being linked to a specific marker and the probability of that linkage being present in any given sib pair. This allows for the possibility that a QTL might be linked in some families but not others.
  • Likelihood-Ratio Statistic (HET-LRT): They use a likelihood-ratio statistic to test for linkage. This statistic compares the likelihood of the data under a model that allows for heterogeneity (where the QTL might be present in some families) versus a model that assumes no linkage or a homogeneous linkage.
  • Sib-Pair Analysis: The analysis focuses on sib pairs (siblings) because they share a predictable proportion of their genes. By comparing how trait values correlate between siblings based on their shared genetic material (IBD - identity by descent), they can infer the presence of linked genes.
  • Data Transformation: To handle the non-normal distribution of blood pressure measurements, they transformed the data using a "normal copula model" to ensure it followed a standard normal distribution. This is a standard practice to improve the performance of statistical models that assume normality.
  • Grid Search for MLEs: Maximum Likelihood Estimates (MLEs) for the model parameters (like the probability of linkage and the effect size) were found using a two-dimensional grid search with a step size of 0.01.
  • Simulation Study: To understand the statistical properties of their new test statistic (HET-LRT), they conducted a simulation study with 10,000 replicates. This helps establish critical values for significance testing.

Datasets

The primary dataset used in this study is the Framingham Heart Study.

  • Framingham Heart Study: This is a well-established, population-based longitudinal study that has collected data on cardiovascular disease risk factors for decades.
  • Study Cohorts: The study includes two cohorts: the original cohort (recruited in 1948) and a second cohort of their offspring and spouses.
  • Data Used: For this analysis, they utilized data from 330 pedigrees, including genotyped markers and phenotyped blood pressure measurements for 1702 subjects. They specifically focused on 1909 subjects who had never been treated for hypertension to avoid confounding effects of medication.
  • Marker Data: A total of 400 genetic markers were used in the genome scan.

Key Insights and Ideas

  • Locus Heterogeneity is Crucial for Complex Traits: The paper strongly suggests that ignoring locus heterogeneity can lead to missed signals when mapping genes for complex traits. Their method, by accounting for this, potentially increases the power to detect true associations.
  • Novel Genetic Regions for Blood Pressure: The identification of two markers (GATA89G08 and GATA23D06) that show linkage but don't overlap with previously reported signals is a significant finding. This suggests they may have uncovered new genetic influences on blood pressure.
  • Validation of Existing Findings: The overlap of two other markers (GATA14E09 and 049xd2) with previously identified linkage regions provides some validation for their method and reinforces the importance of those regions in BP regulation.
  • HET-LRT vs. Other Methods: The paper compares their novel HET-LRT statistic with the standard Haseman-Elston (H-E) method and a homogeneous linkage model (HOM-LRT). While H-E identified more signals, HET-LRT is specifically designed to handle heterogeneity, making its significant findings more robust in that context.

Key Results

  • Four Significant Markers: At a significance level of 0.01, their genome-wide scan identified four markers showing evidence of linkage to blood pressure: GATA89G08, GATA23D06, GATA14E09, and 049xd2.
  • Novelty: Two of these markers (GATA89G08 and GATA23D06) did not show linkage in previous studies, suggesting potential new discoveries.
  • Confirmation: Two markers (GATA14E09 and 049xd2) showed linkage in regions previously implicated in blood pressure regulation.
  • Comparison of Statistics: The Haseman-Elston (H-E) method identified the most significant markers, followed by the homogeneous linkage model (HOM-LRT), and then their novel heterogeneity model (HET-LRT). However, the authors emphasize that HET-LRT is specifically designed for heterogeneity.

Limitations and Future Work

  • Exclusion of Treated Individuals: The analysis excluded individuals on antihypertensive medication. This is a limitation because it means the findings might not directly generalize to the entire population, including those being treated. Careful consideration is needed when applying these results to broader populations.
  • Sib-Pair Data Only: The proposed model is currently designed for sib-pair data. The authors suggest it would be valuable to generalize this model to handle larger, more complex pedigrees.
  • Performance on Selected Samples: The paper notes that the performance of their method on "selected samples" (as opposed to a population-based sample) is unknown and warrants further investigation.
  • Generalizing to General Pedigrees: A key future direction is to extend the model to analyze data from general pedigrees, which would significantly broaden its applicability.
  • Bayesian Approach: The authors mention that they are working on adapting their previous work on Bayesian linkage analysis for qualitative traits to this new heterogeneity model for quantitative traits. This suggests a future Bayesian implementation could be developed.

Summary updated

Abstract

We describe a method for mapping quantitative trait loci that allows for locus heterogeneity. A genome-wide linkage analysis of blood pressure was performed using sib-pair data from the Framingham Heart Study. Evidence of linkage was found on four markers (GATA89G08, GATA23D06, GATA14E09, and 049xd2) at a significance level of 0.01. Two of them (GATA14E09 and 049xd2) seem to overlap with linkage signals reported previously, while the other two are not linked to any known signals.
BibTeX
@inproceedings{yang2003genome,
  author = {Xinqun Yang and Kai Wang and Jian Huang and Veronica J. Vieland},
  title = {{Genome-wide linkage analysis of blood pressure under locus heterogeneity}},
  booktitle = {BMC},
  year = {2003},
  doi = {10.1186/1471-2156-4-s1-s78},
}
Figure