TL;DR
This paper introduces CancerSplicingQTL, a novel database that systematically identifies and catalogs splicing quantitative trait loci (sQTLs) across 33 human cancer types. By integrating vast amounts of genomic and transcriptomic data, this resource aims to bridge the knowledge gap between genetic variants and their functional impact on alternative splicing in cancer, paving the way for a deeper understanding of cancer development and potential therapeutic targets.
What Problem Does This Paper Address?
Alternative splicing (AS) is a crucial process that generates diverse protein isoforms from a single gene, and its dysregulation is a hallmark of cancer, contributing to initiation, progression, and therapy resistance. While genetic variants (SNPs) are known to influence AS events, leading to splicing quantitative trait loci (sQTLs), a comprehensive resource for identifying and studying these sQTLs specifically within the context of human cancers has been lacking. Previous studies have focused on specific tissues or smaller sample sizes, leaving a significant gap in our understanding of how germline genetic variation impacts splicing across the diverse landscape of human cancers. This paper addresses this gap by creating a centralized, user-friendly database of sQTLs derived from a large-scale analysis of The Cancer Genome Atlas (TCGA) data.
Methods
The researchers developed a computational pipeline to identify sQTLs by integrating genotype data with alternative splicing information.
- Data Collection: They utilized genotype data from 9,026 tumor samples across 33 cancer types from The Cancer Genome Atlas (TCGA). Splicing event data, specifically Percent Spliced In (PSI) values, were obtained from TCGASpliceSeq.
- Genotype Processing: Raw genotype data underwent imputation using IMPUTE2 with the 1000 Genomes Phase 3 reference panel to increase SNP density. Rigorous quality control filtering was applied to exclude low-quality SNPs, including those with poor imputation scores, low minor allele frequencies (MAF), high missing rates, or deviations from Hardy-Weinberg equilibrium.
- Splicing Data Processing: PSI values for six types of AS events (skipped exon, retained intron, alternative donor/acceptor sites, alternate terminator, and alternate promoter) were collected. Probes with missing PSI values or located on sex chromosomes were filtered out. To mitigate the impact of outliers, PSI values were normalized to a standard normal distribution based on rank.
- Covariate Inclusion: To account for potential confounding factors, principal components (PCs) derived from EIGENSOFT were used to control for population structure and ethnicity. Additionally, PEER software was employed to infer hidden batch effects and other confounders in the AS data, with the top 15 PEER factors included as covariates. Age, sex, and tumor stage were also incorporated as covariates.
- sQTL Identification: Linear regression was performed using MatrixEQTL to evaluate the association between each SNP and AS events within a ±100 kb window. SNPs with a false discovery rate (FDR) < 0.05 were identified as sQTLs.
- Survival Association Analysis: To identify sQTLs potentially impacting patient outcomes, associations between identified sQTLs and patient overall survival times were assessed using log-rank tests.
- GWAS Integration: To link sQTLs to known disease-associated genetic loci, the identified sQTLs were overlapped with known GWAS risk tag SNPs and their corresponding linkage disequilibrium (LD) regions obtained from the NHGRI GWAS Catalog and SNAP.
Datasets
- The Cancer Genome Atlas (TCGA): Genotype data (level 2) from 9,026 tumor samples across 33 cancer types and corresponding PSI values from TCGASpliceSeq were utilized.
- 1000 Genomes Project Phase 3: Used as a reference panel for genotype imputation.
- NHGRI GWAS Catalog: Used to identify known GWAS risk tag SNPs.
- SNAP: Used to obtain linkage disequilibrium (LD) regions for GWAS tag SNPs.
No new benchmark datasets were introduced; rather, existing large-scale datasets were integrated and analyzed.
Key Insights and Ideas
- Bridging the Gap: The paper successfully bridges a critical gap by providing a comprehensive resource for sQTLs specifically in cancer, moving beyond tissue-specific analyses.
- Functional Relevance of SNPs: By identifying sQTLs, the study offers a mechanistic link between genetic variants (often identified in GWAS) and their functional impact on gene expression through alternative splicing, which is crucial for understanding disease susceptibility and progression.
- Survival Association: The identification of sQTLs associated with patient survival times highlights their potential clinical relevance and suggests that genetic variations influencing splicing could directly impact cancer prognosis.
- GWAS Integration for Causal Variant Discovery: Linking sQTLs to GWAS loci provides a powerful approach to prioritize candidate causal variants within GWAS risk regions, offering a more refined understanding of disease-associated genetic signals.
- Database Utility: The creation of a user-friendly database (CancerSplicingQTL) makes this extensive data accessible for researchers, facilitating further investigation into the role of splicing in cancer.
Key Results
- Vast sQTL Collection: The study identified a total of 4,599,598 sQTLs across all 33 cancer types.
- Survival-Associated sQTLs: 17,072 sQTLs were found to be significantly associated with patient overall survival times.
- GWAS-Related sQTLs: 1,180,132 sQTLs overlapped with known GWAS linkage disequilibrium regions, providing potential functional interpretations for GWAS findings.
- Cancer Type Variation: The number of identified sQTLs varied significantly across cancer types, ranging from 64 in cholangiocarcinoma (CHOL) to 574,577 in thyroid carcinoma (THCA).
- sQTL Distribution: Most sQTLs were located within a ±31 kb region flanking the affected AS events, suggesting a strong cis-regulatory effect.
- AS Event Types: Alternate terminator (AT), exon skip (ES), and alternate promoter (AP) were the most frequently affected AS event types by sQTLs.
Limitations and Future Work
- Limitations:
- The study acknowledges that the identification of causal variants remains a challenge, and the identified sQTLs are candidates that require further validation.
- While the database is comprehensive, the authors note that the number of sQTLs can be influenced by sample size, and some cancer types have limited sample availability.
- The paper mentions that batch effect correction can lead to the loss or gain of sQTLs, indicating potential variability in the results depending on data processing steps.
- Future Work:
- The authors plan to continue updating CancerSplicingQTL with more cancer samples and genetic/splicing information to maintain its utility.
- They aim to expand the database to include more genetic and splicing-related information, further enhancing its comprehensiveness.
- The database is intended to serve as a valuable resource for bridging the knowledge gap from sequence variants to phenotypes in cancer genetics, facilitating further research into personalized medicine and therapeutic strategies.
Summary updated