Region-Based Analyses of Existing Genome-Wide Association Studies Identifies Novel Potential Genetic Susceptibility Regions for Glioma

Karen Alpen, Robert J. Maclnnis, Claire M. Vajdic, John Lai, James G. Dowty, Eng‐Siew Koh, Elizabeth Hovey, Rosemary Harrup, Tuong L. Nguyen, Shuai Li, David Joseph, Geza Benke, Pierre‐Antoine Dugué, Melissa C. Southey, Graham G. Giles, Anna K. Nowak, Katharine J. Drummond, Daniel F. Schmidt, John L. Hopper, Miroslaw K. Kapuscinski, Enes MakalicCancer 2024 · DOI

TL;DR

This study leverages novel region-based analysis methods on existing genome-wide association study (GWAS) data to identify new genetic regions associated with glioma risk. By looking at groups of genetic markers rather than individual ones, the researchers uncovered 11 potential new susceptibility regions, with two showing strong associations that differ between males and females, suggesting sex-specific genetic influences on glioma.

What Problem Does This Paper Address?

Glioma, a devastating brain cancer, has a low survival rate, and despite decades of research, a significant portion of its genetic susceptibility remains unexplained. Traditional GWAS, which analyze single genetic markers (SNPs) independently, often require very large sample sizes to detect associations due to stringent statistical thresholds needed to correct for the millions of tests performed. This is particularly challenging for rare diseases like glioma. While 34 genetic susceptibility regions have been identified, these account for less than 40% of the estimated familial risk, leaving a substantial gap in our understanding of the genetic underpinnings of the disease. This paper aims to bridge this gap by employing advanced region-based analytical approaches to re-examine existing GWAS data, hoping to uncover novel susceptibility regions that might have been missed by conventional methods.

Methods

The study employed a two-stage approach for identifying novel susceptibility regions.

  1. Hypothesis Generation (Stage 1): A machine learning algorithm called DEPTH (DEPendency of association on the number of Top Hits) was used. DEPTH analyzes groups of contiguous SNPs within sliding windows across the genome. It's a hypothesis-generating tool that doesn't require strict statistical correction at this stage and can model non-linear relationships between SNPs. This stage aimed to prioritize candidate regions by identifying those with a significant association signal across multiple datasets.
  2. Association Testing and Validation (Stages 2-4):
    • Region-Based Association Testing (Stage 2): The generalized Berk-Jones (GBJ) statistic was used. GBJ is a set-based statistical method that aggregates weak association signals from multiple SNPs within a predefined region or gene, offering better power to detect associations compared to single-SNP analyses, especially when signals are sparse. This stage tested the candidate regions identified by DEPTH for association with glioma risk, adjusting for covariates like sex and age.
  3. Validation (Stage 3): Candidate regions were independently validated using summary statistics from a separate study (UCSF/Mayo).
  4. Meta-Analysis (Stage 4): A meta-analysis combining data from the discovery studies (GliomaScan, AGOG, GICC) and the validation study (UCSF/Mayo) was performed to increase statistical power and confirm findings.

The study also performed sex-specific analyses throughout these stages, recognizing potential sex differences in glioma susceptibility.

Datasets

The study utilized data from three independent glioma case-control studies for discovery and analysis:

  • GliomaScan: 1,316 cases / 1,293 controls
  • Australian Genomics and Clinical Outcomes of Glioma Consortium (AGOG): 560 cases / 2,237 controls
  • Glioma International Case-Control Study (GICC): 4,000 cases / 2,411 controls

For independent validation, summary statistics from the:

  • UCSF/Mayo Clinic study: 2,141 cases / 1,889 controls

The 1000 Genomes Project European reference panel was used for linkage disequilibrium (LD) estimation during the validation and meta-analysis stages.

Key Insights and Ideas

  • Region-based analysis is powerful for rare diseases: By aggregating signals from multiple SNPs, this approach can detect associations that individual SNPs might miss, which is crucial for diseases with complex genetic architectures and smaller sample sizes.
  • Sex matters in glioma genetics: The identification of distinct genetic associations for male and female glioma risk highlights that sex is not just a demographic factor but a biological one that influences genetic susceptibility. This underscores the need for sex-specific analyses in future genetic studies of glioma.
  • Synaptic genes as potential players: The discovery of regions containing neurotransmitter receptor genes suggests that genes involved in neuronal communication and synaptic function might play a role in glioma susceptibility, opening up new avenues for research into the biological mechanisms of the disease.
  • Novelty in known regions: The study identified regions that were not previously linked to glioma risk, suggesting that our understanding of the genetic landscape of glioma is still incomplete.

Key Results

  • The study identified 11 novel candidate genomic regions associated with glioma risk.
  • Two of these regions, 16p13.3 (containing RBFOX1) and 1p36.21 (containing PRDM2), showed significant associations in the meta-analysis.
  • 16p13.3 was significantly associated with female glioma risk.
  • 1p36.21 was significantly associated with male glioma risk.
  • Three other regions contained neurotransmitter receptor genes: 7q31.33 (GRM8), 5q35.2 (DRD1), and 15q13.3 (CHRNA7), suggesting a potential role for synaptic genes.
  • A specific SNP (rs2346609) in the RBFOX1 gene was strongly associated with female glioma susceptibility, while another SNP (rs35042965) in the PRDM2 gene showed a protective effect against male glioma risk.

Limitations and Future Work

The authors acknowledge several limitations:

  • Glioma Subtype Data: The lack of detailed glioma subtype information for the GICC study and the small sample sizes for astrocytoma and oligodendroglioma in AGOG and GliomaScan limited the discovery of subtype-specific susceptibility regions.
  • Threshold for DEPTH: The study focused on regions with a DEPTH PLO score of ≥2.0, meaning regions with weaker but still potentially significant signals (PLO between 1.0 and 2.0) might have been missed.
  • Potential False Negatives: Some candidate regions identified by DEPTH might not have met GBJ significance thresholds due to DEPTH's ability to detect complex SNP interactions that linear models like GBJ might miss.
  • Data Overlap: The UCSF/Mayo validation dataset had some overlapping controls with the GICC study, meaning the validation was not entirely independent.
  • Proxy for GBM/non-GBM: For GBM and non-GBM analyses in the UCSF/Mayo data, IDHwt and IDH mutant status were used as proxies, which might not perfectly align with the GBM/non-GBM classifications used in the discovery datasets.
  • Ancestry Restriction: The findings are primarily applicable to individuals of European ancestry.

Future work should focus on further investigating the 11 identified regions, particularly 1p36.21 and 16p13.3, through larger, sex-specific studies. Exploring the role of synaptic genes in glioma susceptibility is also a promising direction. Expanding these analyses to diverse ancestral populations is crucial for a comprehensive understanding of glioma genetics.

Summary updated

Abstract

Further investigation of the potential susceptibility regions identified in our study may lead to a better understanding of glioma genetic risk and the underlying biological etiology of glioma. Our study suggests sex may play a role in genetic susceptibility and highlights the importance of sex-specific analysis in future glioma research.
BibTeX
@inproceedings{alpen2024region,
  author = {Karen Alpen and Robert J. Maclnnis and Claire M. Vajdic and John Lai and James G. Dowty and Eng‐Siew Koh and Elizabeth Hovey and Rosemary Harrup and Tuong L. Nguyen and Shuai Li and David Joseph and Geza Benke and Pierre‐Antoine Dugué and Melissa C. Southey and Graham G. Giles and Anna K. Nowak and Katharine J. Drummond and Daniel F. Schmidt and John L. Hopper and Miroslaw K. Kapuscinski and Enes Makalic},
  title = {{Region-Based Analyses of Existing Genome-Wide Association Studies Identifies Novel Potential Genetic Susceptibility Regions for Glioma}},
  booktitle = {Cancer},
  year = {2024},
  doi = {10.1158/2767-9764.crc-24-0385},
}
Figure