Validating reference-based algorithms to determine cell-type heterogeneity in ovarian cancer DNA methylation studies

Area of Science:

• Epigenetics
• Cancer Biology
• Bioinformatics

Background:

• Accurate cell composition analysis in cancer tissues is vital for biomarker discovery and prognosis.
• Ovarian cancer deconvolution studies face challenges due to genetic mutations and rearrangements.

Purpose of the Study:

• To optimize and validate a robust DNA methylation-based protocol for cell type deconvolution in ovarian cancer samples.
• To compare the performance of state-of-the-art deconvolution methods, focusing on the impact of reference panel size.

Main Methods:

• Optimization of a DNA methylation-based protocol for deconvolution.
• Comparison of HEpiDISH, MethylCIBERSORT, and ARIC deconvolution algorithms.
• Validation using in-silico mixtures, external datasets, and a cohort of 72 ovarian disease patients.

Main Results:

• A robust DNA methylation-based protocol using MethylCIBERSORT was implemented and validated.
• A large reference panel (247 samples) was found to enhance deconvolution robustness.
• The validated protocol demonstrated reliable performance across multiple ovarian cancer cohorts.

Conclusions:

• A refined, reference-based algorithm was developed for accurate cell type composition analysis in ovarian cancer.
• This validated protocol is suitable for large-scale cancer biology studies.
• The findings support the use of DNA methylation deconvolution for improved understanding of ovarian cancer heterogeneity.