Identifying regulatory relationships among genomic loci, biological pathways, and disease

Jung Hoon Woo, Sung-Bum Cho, Eunjee Lee, Ju Han KimArtif. Intell. Medicine 2010 · DOI

TL;DR

This paper introduces a novel computational approach, Differential Allelic Co-expression (DACE), to uncover complex regulatory relationships between genetic variations (SNPs) and entire biological pathways, not just individual genes. This integrative method is crucial for understanding how genetic predispositions translate into complex diseases by revealing how SNPs influence the coordinated activity of multiple genes within pathways and, ultimately, disease outcomes.

What Problem Does This Paper Address?

Understanding the genetic basis of complex diseases is a major challenge. Traditional genetic studies often focus on single genes or simple associations between genetic variants and traits. However, complex diseases arise from intricate interactions among multiple genes and pathways. Existing methods, like single-gene association analysis, fail to capture how genetic variations might influence the coordinated behavior of multiple genes within a pathway, which is essential for understanding disease mechanisms. This paper aims to bridge this gap by developing a method that can identify regulatory loci affecting not only individual gene expression but also the inter-correlations between genes within a pathway, thereby providing a more holistic view of genetic influence on disease.

Methods

The authors propose and apply the Differential Allelic Co-expression (DACE) test. This method goes beyond traditional single-gene association tests by examining how a SNP's genotype influences the correlation structure among the expression levels of genes within a defined pathway or gene set.

Here's a breakdown:

  • Core Idea: DACE tests if a SNP's genotype leads to significant differences in how genes within a pathway co-express (i.e., their expression levels move together).
  • Process:
    1. Grouping by Genotype: Samples are grouped based on their genotype for a specific SNP.
  • Correlation Calculation: Within each genotype group, the Pearson correlation coefficients between the expression levels of all gene pairs in a pathway are calculated.
  • Fisher's z-transformation: This is applied to the correlation coefficients to make them more amenable to statistical testing, as they are not normally distributed.
  • Linear Model: A linear model framework is used to test whether the SNP genotype has a significant effect on these correlation coefficients, indicating a regulatory influence on the pathway's co-expression pattern.
  • Comparison: The DACE test is compared against:
  • Principal Component Analysis (PCA): Used to reduce the dimensionality of gene expression data within a pathway, with the first principal component (PC1) treated as a representative trait. SNPs associated with PC1 are considered regulatory.
  • Single-gene association analysis: Standard regression analysis testing the association between a single SNP and the expression level of a single gene.

Datasets

The study utilizes two primary datasets:

  1. Lymphoblastoid cell dataset: This dataset, from the Genetic Analysis Workshop (GAW) 15, consists of genotype data for 2882 genome-wide SNPs and mRNA expression data for 8793 genes from 56 independent individuals. The data was normalized using robust multichip average (RMA).
  2. Human glioblastoma dataset: This dataset was obtained from the Broad Institute website and includes SNP and mRNA expression data for 34 glioblastoma patients. The platforms used were Affymetrix 100K SNP chip and U133A for mRNA expression. Approximately 52,885 SNP probes were used after filtering.

Key Insights and Ideas

  • Beyond Single Genes: The DACE test highlights that genetic regulation of complex diseases often involves coordinated changes in multiple genes within pathways, not just isolated gene effects.
  • Capturing Inter-gene Relationships: The core innovation is the focus on how SNPs alter the relationships (correlations) between genes, which is a more nuanced view than just affecting individual expression levels.
  • Biological Explanation for SNPs: The integrative approach helps provide biological context for SNPs, especially those located in non-coding regions (like introns), by linking them to functional pathways and disease outcomes.
  • DACE Outperforms PCA: The study suggests that DACE is more sensitive than PCA in identifying regulatory associations, particularly for subtle effects on co-expression patterns.
  • Linking Genotype to Survival: The glioblastoma analysis demonstrates a direct link between specific SNP genotypes, pathway co-expression patterns, and patient survival, offering a potential mechanism for how genetic variation influences disease prognosis.

Key Results

  • Lymphoblastoid Data:
    • PCA identified associations with six pathways, often correlating with the major contributing gene within that pathway.
  • DACE identified associations not detected by single-gene analysis. For instance, SNP rs1294028 was significantly associated with the inflammatory response pathway via DACE, showing altered correlation structures among genes, even though no individual gene in the pathway showed a significant association with this SNP.
  • Seven immune system-related pathways showed significant SNP associations using DACE.
  • Glioblastoma Data:
    • DACE identified 4582 SNP-pathway pairs associated with six different pathways.
  • Of these, 231 SNPs showed a significant correlation with patient survival length (P < 0.05).
  • A specific SNP, rs6661074, was found to be strongly associated with both the Peptide GPCRs pathway and glioblastoma patient survival, demonstrating a co-localization of SNP-pathway and SNP-disease associations. Patients with the GG genotype at rs6661074 had better survival and lower intra-correlation among genes in the Peptide GPCRs pathway.

Limitations and Future Work

  • Limitations:
    • The study acknowledges that the pattern of variation in PCA can be biased by the gene with the largest contribution to a principal component, potentially masking other regulatory effects.
  • The biological explanation for SNPs in non-coding regions (like introns) can still be challenging, even with pathway-level analysis.
  • Future Work:
    • The authors suggest that their integrative approach, using set-wise genetical genomics to identify SNP/pathway associations, can provide more relevant explanations for complex disease studies than traditional methods.
  • The findings suggest that further investigation into the role of specific SNPs and pathways in disease progression and prognosis is warranted. The DACE test is presented as a promising tool for future research in this area.

Summary updated

BibTeX
@article{woo2010identifying,
  author = {Jung Hoon Woo and Sung-Bum Cho and Eunjee Lee and Ju Han Kim},
  title = {{Identifying regulatory relationships among genomic loci, biological pathways, and disease}},
  journal = {Artif. Intell. Medicine},
  year = {2010},
  doi = {10.1016/j.artmed.2010.03.001},
}
Figure