Abstract
This report summarizes a series of corrigenda published in Nature Genetics (2014) addressing significant errors in high-profile genomic studies. The corrections focus on the impact of population stratification and improper penetrance estimates on the validity of Genome-Wide Association Studies (GWAS). Key findings include the retraction of a major susceptibility locus for esophageal squamous cell carcinoma (ESCC) and the provision of corrected mutation rates for maize inbred lines.
Introduction: The Fragility of Genomic Associations
The primary challenge in genomic research is ensuring the reproducibility and accuracy of GWAS. Two common pitfalls often lead to false-positive results: improper penetrance estimates and population stratification. Inadequate control for these factors can result in overestimating the disease risk of specific mutations or misidentifying ancestral genetic noise as disease-linked variants.
Methodological Improvements
To address these issues, researchers re-analyzed their data using more stringent statistical controls:
- Genetic Matching: For ESCC studies, researchers transitioned from simple case-control comparisons to "genetically matched subsets," pairing each case with a control of nearly identical ancestral lineage to eliminate noise.
- Penetrance Recalculation: Models were adjusted to reflect real-world disease prevalence (e.g., 5.3% for developmental delay) rather than artificial study ratios.
- Independent Replication: Significant SNPs were tested against independent datasets, such as the National Cancer Institute (NCI) and Beijing ESCC GWAS, to verify if signals persisted.
The "Phantom" ESCC Locus
The most significant outcome of this re-analysis was the retraction of the association for SNP rs13042395 at the C20orf54 locus. Originally reported as a highly significant hit ($P = 1.51 \times 10^{-11}$), it was revealed to be a statistical artifact of population stratification or chance. Upon re-analysis of the Henan data, the Odds Ratio (OR) shifted from a significant 0.66 to a non-significant 0.88 ($P = 0.09$). Three other SNPs (rs17761864, rs6503659, and rs2847281) were also found to lack supporting evidence once ancestry was properly accounted for.
Corrected Genomic Metrics in Maize
In addition to human studies, the corrigenda provided corrected baseline mutation rates and rare allele distributions for four related maize inbred lines (B73, 8112, 478, and Zheng58).
The analysis of rare alleles across the whole genome, genic regions, and coding regions (Figure 1) highlights the genomic variation within these lines. Furthermore, corrected mutation rates for single-base substitutions, indels, and copy number variations (CNVs) were established across specific lineage transitions.
As shown in Table 1 (Figure 2), single-base substitution rates were refined to ranges between $9.92 \times 10^{-9}$ and $2.02 \times 10^{-8}$ per site per year. These corrections are vital for accurate evolutionary and functional genomic modeling in maize.
Conclusion
These corrigenda serve as a critical reminder of the necessity for rigorous ancestry control and mandatory replication in genomic research. Findings from single cohorts should be treated as preliminary until validated in independent, ancestrally diverse populations to avoid the propagation of false-positive associations in the scientific literature.
Summary updated