Simultaneous statistical inference for epigenetic data.

Area of Science:

• Epigenetics
• Statistical Genomics
• Bioinformatics

Background:

• Epigenetic research generates complex data, often with unverifiable parametric distribution assumptions.
• Simultaneous analysis of multiple genetic loci and interrelated biological parameters is common.
• Existing statistical methods struggle with unknown dependency structures and require robust multiple testing correction.

Purpose of the Study:

• To develop a nonparametric statistical framework for two-group comparisons of epigenetic data.
• To address the challenges of multiple testing and unknown dependency structures in epigenetic analyses.
• To provide a reliable method for detecting group differences in epigenetic data at specific loci and parameters.

Main Methods:

• Introduction of a nonparametric statistical framework for comparing two samples with independent multivariate observables.
• Adaptation of multivariate multiple permutation tests to handle families of hypotheses regarding relative effects.
• Application of the closure principle to control the family-wise error rate for simultaneous locus/parameter-specific null hypotheses.

Main Results:

• The multivariate multiple permutation test effectively maintains the pre-assigned type I error level for the global null hypothesis.
• The proposed methodology successfully controls the family-wise error rate for simultaneous testing of multiple hypotheses.
• Demonstrated reliable detection of group differences in epigenetic data through practical applications.

Conclusions:

• The developed nonparametric framework offers a robust solution for analyzing complex epigenetic data.
• The methodology ensures valid conclusions for specific loci and parameters by controlling error rates.
• This approach enhances the reliability of detecting epigenetic group differences in high-dimensional biological data.