Cross-cohort DNA methylation analysis shows CpG island promoter hypermethylation features associated with cervical lesion grade.
TL;DR
The reanalysis supports reproducible CpG island and promoter hypermethylation features associated with cervical lesion grade, but processed data, cohort heterogeneity, cross-sectional tumor associations, and absence of prospective validation limit interpretation to candidate atlas generation.
Abstract
Background
Cervical lesion progression is accompanied by epigenomic alterations, but reproducible methylation features across public cohorts require cautious re-analysis. STUDY
Design
AND
Methods
We performed a secondary analysis of public Illumina 450K methylation cohorts: GSE46306 as discovery (20 normal, 17 CIN3, six carcinoma), GSE99511 and GSE186835 as validation, and TCGA-CESC tumors for methylation-RNA expression association. Shared autosomal CpGs were screened for monotonic grade-associated change, cross-cohort directional consistency, promoter annotation, regional aggregation, sample-level score performance, and tumor expression correlation. No intervention, dose, treatment regimen, blinding, or safety outcome was involved.
Results
Among 340, 248 shared autosomal CpGs, 3, 008 met robust cross-cohort criteria, including 2, 991 hypermethylated and 17 hypomethylated sites. Robust sites mapped to 628 promoter genes and 660 multi-CpG DMR-like regions. A fixed 200-CpG score yielded AUCs of 0.733-0.963 for CIN3 or worse versus controls and 0.971-0.988 for cancer versus non-cancer. In TCGA-CESC, 394 of 610 detectable robust promoter genes showed significant negative methylation-expression correlations.
Conclusions
The reanalysis supports reproducible CpG island and promoter hypermethylation features associated with cervical lesion grade, but processed data, cohort heterogeneity, cross-sectional tumor associations, and absence of prospective validation limit interpretation to candidate atlas generation.