A Network-guided Association Mapping Approach from DNA Methylation to Disease

Area of Science:

• Genomics
• Bioinformatics
• Cancer Research

Background:

• Aberrant DNA methylation is implicated in cancer development.
• Understanding DNA methylation-cancer associations is challenging due to complex mechanisms and limited sample sizes.
• Advancements in data availability (DNA methylation, gene expression, disease status) offer new opportunities for computational approaches.

Purpose of the Study:

• To propose a novel network-guided association mapping approach from DNA methylation to disease (NAMDD).
• To improve the identification of significant DNA methylation sites associated with cancer.
• To investigate underlying mechanisms linking DNA methylation and disease through data integration.

Main Methods:

• Developed NAMDD, a machine learning approach integrating multiple data types (DNA methylation, gene expression, disease status).
• Employed a stability selection strategy to enhance the robustness of identified associations.
• Utilized network-guided analysis to map methylation-disease paths.

Main Results:

• NAMDD outperforms existing methods (NsRRR, PCLOGIT) in identifying disease-related methylation sites, particularly under false positive control.
• Experimental results on synthetic and real ovarian cancer data demonstrate NAMDD's superior performance.
• Identified significant path associations in ovarian cancer data, providing potential biological insights.

Conclusions:

• NAMDD is an effective tool for uncovering DNA methylation-disease associations.
• The method enhances the quality and reliability of identified methylation sites.
• Findings from ovarian cancer data suggest potential novel biological pathways involved in cancer development.