Background
Common methods of identification of differentially methylated genes (DMGs) mainly detect differences between case and control groups, which cannot tell whether a gene is differentially methylated in a specific disease sample (first scenario), and are not applicable for the study with no normal control (one-phenotype, second scenario). Also, these methods have low detection capacity at the control-limited (third) scenario.
Results
we developed a method, termed RankDMG, to analyze DNA methylation data in the three special scenarios. For the individualized DMG analysis, RankDMG showed remarkable performances in simulated and real data, independent of measured platforms. Using DMGs detected by common methods as ‘gold standard’, the DMGs identified by RankDMG using only one-phenotype data were comparable to those detected by common methods using case-control samples. Moreover, even when the number of disease samples reduced to five, RankDMG could also identify disease-related DMGs for control-limited data.
Conclusion
RankDMG provides a novel tool to dissect the inter-individual heterogeneity of tumor at epigenetic level, and it could analyze the one-phenotype and control-limited methylation data. RankDMG is provided as an open source tool via https://github.com/FunMoy/RankDMG.