TL;DR
This study reveals that a surprising number of genetic variants are found in both our permanent germline DNA and our transient somatic (body) cells, suggesting shared underlying mutation processes. By analyzing these shared variants, the researchers not only gain insights into the fundamental mechanisms of mutation but also highlight the importance of robust germline filtering for accurate somatic variant detection and protecting individual privacy.
What Problem Does This Paper Address?
Understanding the origins and patterns of genetic mutations is crucial for both diagnosing diseases and comprehending human evolution. Mutations can occur in two main ways: in the germline (sperm and egg cells), which are heritable, or in somatic cells, which are not. While these settings have different influences (e.g., environmental exposures for somatic cells, internal cellular processes for germline), recent research suggests their mutation rates are surprisingly similar.
A key challenge is distinguishing true disease-causing variants from the vast background of neutral or near-neutral mutations. To do this effectively, researchers need accurate models of "background" mutation rates. Combining data from both somatic and germline mutations could create more powerful models, but this requires understanding how similar their mutation patterns truly are. Furthermore, accurately identifying somatic mutations relies on distinguishing them from any germline DNA that might contaminate a somatic sample, which is critical for both research accuracy and the privacy of individuals whose samples are used.
Methods
The researchers employed a statistical framework to quantify the overlap between somatic and germline single nucleotide variants (SNVs), termed "co-occurring SNVs" (cSNVs). They compared the observed number of cSNVs to what would be expected by chance (statistical independence). To explain the observed excess of cSNVs, they developed a "partitional dependence" metric to assess how well different factors, such as nucleotide context and genomic region, could account for this overlap.
Simulations were used to benchmark the power of their SNV recurrence approach and to understand how hypothetical mutational processes affecting both somatic and germline DNA could lead to excess cSNVs. They also utilized a read depth analysis on matched normal samples to estimate the extent of germline contamination in somatic samples.
Datasets
- gnomAD (genome Aggregation Database): A large database of germline whole-exome sequencing data from approximately 120,000 individuals.
- TCGA (The Cancer Genome Atlas): A dataset of somatic whole-exome sequencing data from approximately 10,000 cancer patients, harmonized across cancer types.
- denovo-db: A compendium of human de novo germline variants confirmed through parental-offspring trio sequencing.
- Normal somatic stem cell data: Used to assess cSNV enrichment in non-cancerous tissues.
Key Insights and Ideas
- Shared Mutation Processes: The significant overlap between somatic and germline variants suggests that fundamental mutational processes are shared between these two distinct cellular environments, despite their different influences.
- Nucleotide Context is King: The primary driver of this shared mutational landscape appears to be the nucleotide context surrounding a mutation, particularly the N[C->T]G motif, which is known to be highly mutable in both soma and germline.
- Signature 1's Ubiquity: The endogenous mutational process known as "Signature 1" (spontaneous deamination of 5-methylcytosine) is a major contributor to this shared mutational pattern, being active in both somatic and germline DNA.
- Germline Contamination is a Real Concern: The study highlights that germline contamination in somatic samples is more prevalent than previously estimated and can inflate the observed rate of shared variants. This has implications for both data accuracy and the privacy of individuals.
- Predictive Power of Mutation Models: The fact that known determinants of mutation rates can explain a large portion of the observed shared variants indicates the genomics community has made significant progress in understanding these processes.
Key Results
- Three Times More Shared Variants: The study found approximately three times more SNVs shared between somatic and germline call-sets than would be expected by chance.
- 97% Explained by Known Factors: Over 90% of this excess overlap can be explained by correlated mutation rates of basic nucleotide contexts in the soma and germline. Specifically, nucleotide context alone explained 97.2% of the excess recurrence.
- Germline Contamination Estimate: The study estimates approximately 1 leaked germline variant per strictly-filtered somatic exome, accounting for about 4% of the observed overlapping variants.
- Read Depth Matters for Privacy: Increased read depth in matched normal samples is associated with a decreased cSNV rate, suggesting that a read depth of 100 or more is necessary to minimize germline contamination and protect genomic privacy.
- Cancer Type Variation: The degree of overlap varies significantly across different cancer types, suggesting that biological factors and specific mutational sources (like environmental exposures) play a role.
Limitations and Future Work
- Binarization Limitation: The study binarized SNV data (present/absent) to handle differences in allele counting across datasets. This simplification, while necessary, may lead to fewer observed excess cSNVs at highly mutable sites.
- De Novo Variant Interpretation: The higher cSNV rate observed with de novo germline variants compared to gnomAD variants is noted as a potential artifact of gnomAD's large sample size and binarization, requiring further investigation.
- Undiscovered Determinants: While a large portion of shared variants is explained, a small excess remains. This suggests that undiscovered determinants of mutation rates might exist, potentially acting inconsistently across genomic sites or being correlated with known factors.
- Other Variant Types: The study focused on SNVs due to data availability. Future work could explore similar patterns for insertions, deletions, copy number changes, and structural variants.
- Genomic Region Impact: While nucleotide context is a major driver, the impact of genomic region on shared variants was found to be minor, suggesting further investigation into regional mutation rate differences.
- Germline Contamination Nuances: The higher estimate of germline contamination compared to previous studies warrants further exploration, potentially due to differences in filtering strategies or the sensitivity of read-depth analysis.
Summary updated