Pancan-meQTL: a database to systematically evaluate the effects of genetic variants on methylation in human cancer

Jing Gong, Hao Wan, Shufang Mei, Hang Ruan, Zhao Zhang, Chun-Jie Liu, An-Yuan Guo, Lixia Diao, Xiaoping Miao, Leng HanNucleic Acids Res 2019 · DOI

TL;DR

This paper introduces Pancan-meQTL, the first comprehensive database that systematically links genetic variants (SNPs) to DNA methylation patterns across 23 different human cancer types. By integrating vast amounts of data, it provides a crucial resource for understanding how genetic variations influence cancer development and progression through epigenetic mechanisms.

What Problem Does This Paper Address?

Understanding the functional consequences of genetic variations, especially those in non-coding regions of the genome, is a major challenge in deciphering complex diseases like cancer. While Genome-Wide Association Studies (GWAS) have identified numerous genetic loci associated with cancer risk, the underlying molecular mechanisms often remain elusive. One critical mechanism that links genotype to phenotype is DNA methylation, an epigenetic modification that regulates gene expression. Aberrant DNA methylation is a hallmark of cancer. However, a systematic, pan-cancer resource to explore the relationship between genetic variants and DNA methylation (methylation quantitative trait loci, or meQTLs) was lacking. This gap hinders our ability to interpret GWAS findings and understand the interplay between genetics and epigenetics in cancer.

Methods

The authors employed a systematic approach to identify meQTLs across multiple cancer types. They utilized the vast datasets from The Cancer Genome Atlas (TCGA), which includes genome-wide genotype and DNA methylation data for thousands of cancer samples.

  • Data Collection and Preprocessing: Genotype data (level 2) and methylation beta values (level 3) were downloaded from the TCGA data portal. Samples from related cancer types (e.g., colon and rectum adenocarcinoma) were merged, and cancer types with fewer than 100 samples were excluded. Methylation probes on sex chromosomes were removed, and probes were filtered based on missing data, multiple genomic mappings, and the presence of known SNPs. Methylation values were transformed into a standard normal distribution to mitigate outlier effects.
  • Covariate Inclusion: To control for potential confounders, several covariates were included in the analysis: the top five principal components from smartpca (for ethnicity), the first 15 PEER factors (for batch effects), and age, sex, and tumor stage.
  • meQTL Identification: For each cancer type, linear regression was performed using the MatrixEQTL tool to test the association between each SNP and CpG site. meQTLs were defined as SNPs with a false discovery rate (FDR) < 0.05 and an absolute correlation coefficient (|r|) ≥ 0.3. SNPs located within 1 Mb of a probe were classified as cis-meQTLs, while those beyond 1 Mb were classified as trans-meQTLs.
  • Survival Association: meQTLs associated with patient overall survival were identified using the log-rank test.
  • GWAS Overlap: meQTLs were compared against known GWAS loci to identify those overlapping with established cancer risk regions.
  • Enrichment Analysis: To assess the functional relevance of meQTLs, enrichment analyses were performed against transcription factor binding sites (TFBSs) and GWAS loci, using a control dataset of non-meQTL SNPs matched for minor allele frequency and distance.

Datasets

  • The Cancer Genome Atlas (TCGA): This was the primary source of data, providing genome-wide genotype and DNA methylation data for approximately 10,000 cancer samples across 23 cancer types.
  • dbSNP: Used for SNP locations.
  • 1000 Genomes Project: Used for SNP annotation and linkage disequilibrium (LD) analysis.
  • GWAS Catalog: Used to identify risk SNPs from GWAS studies.
  • ENCODE Project: Used for transcription factor binding site (TFBS) data.

No new benchmarks were introduced; rather, existing large-scale datasets were integrated and analyzed.

Key Insights and Ideas

  • Bridging the Genotype-Epigenotype Gap: The core idea is that meQTLs serve as a crucial molecular link between genetic variations (genotype) and epigenetic modifications (DNA methylation), which in turn influence gene expression and ultimately disease phenotypes like cancer.
  • Pan-Cancer Perspective: By analyzing meQTLs across 23 cancer types, the study allows for the identification of both cancer-specific and common epigenetic regulatory mechanisms influenced by genetics. This comparative approach is vital for understanding the heterogeneity and commonalities in cancer development.
  • Functional Interpretation of GWAS Loci: The database provides a powerful tool to interpret the functional role of GWAS-identified variants, many of which reside in non-coding regions. By linking these variants to specific methylation changes, researchers can gain insights into how they contribute to cancer risk.
  • Identification of Survival-Associated Variants: The discovery of meQTLs associated with patient survival highlights their potential as prognostic biomarkers or therapeutic targets.

Key Results

  • Vast meQTL Catalog: The study identified a substantial number of meQTLs: 8,028,964 cis-meQTLs and 965,050 trans-meQTLs across 23 cancer types.
  • Survival Associations: 23,432 meQTLs were found to be significantly associated with patient overall survival times.
  • GWAS Overlap: 2,214,458 meQTLs overlapped with known GWAS loci, suggesting a direct link between genetic predisposition and epigenetic regulation in cancer.
  • Enrichment in Regulatory Elements: meQTLs were found to be significantly enriched in transcription factor binding sites (TFBSs) and GWAS loci, underscoring their functional relevance in gene regulation.
  • Database Availability: The Pancan-meQTL database is publicly accessible via a user-friendly web interface, facilitating browsing, searching, and downloading of data.

Limitations and Future Work

  • Limitations:
    • The study focused on probes from the Illumina HumanMethylation450 BeadChip array, which may not cover all relevant CpG sites.
  • The analysis was limited to samples available in TCGA with both genotype and methylation data, and certain cancer types with smaller sample sizes were excluded.
  • The definition of cis-meQTLs was limited to a 1 Mb window, which might miss some long-range cis-regulatory effects.
  • Future Work:
    • The authors plan to continually update the Pancan-meQTL database as new large-scale genotypic and methylation data become available.
  • The database is intended to be a valuable resource for future research into the roles of genetics and epigenetics in cancer, potentially leading to improved interpretation of genetic risk factors and the identification of new biomarkers and therapeutic targets.

Summary updated

BibTeX
@article{gong2019pancan,
  author = {Jing Gong and Hao Wan and Shufang Mei and Hang Ruan and Zhao Zhang and Chun-Jie Liu and An-Yuan Guo and Lixia Diao and Xiaoping Miao and Leng Han},
  title = {{Pancan-meQTL: a database to systematically evaluate the effects of genetic variants on methylation in human cancer}},
  journal = {Nucleic Acids Res.},
  year = {2019},
  doi = {10.1093/nar/gky814},
}
Figure