Calculating detection limits and uncertainty of reference-based deconvolution of whole-blood DNA methylation data

Area of Science:

• Epigenetics
• Bioinformatics
• Genomics

Background:

• DNA methylation (DNAm)-based cell mixture deconvolution (CMD) is crucial for analyzing heterogeneous tissues in epigenome-wide association studies (EWAS).
• Current CMD methods lack defined detection limits, hindering the reliable identification of low-fraction cell types.
• Quantifying uncertainty in DNAm-based CMD has received limited attention, impacting study reproducibility.

Purpose of the Study:

• To develop analytical frameworks for determining cell-specific limits of detection in DNAm-based CMD.
• To establish methods for quantifying uncertainty in DNAm-based CMD.
• To improve the rigor and replicability of EWAS that utilize CMD.

Main Methods:

• Development of analytical frameworks to calculate cell-specific detection limits.
• Implementation of methods to quantify the uncertainty associated with CMD.
• Application of these frameworks to DNAm data from mixed biological specimens.

Main Results:

• Established protocols for determining the minimum detectable fraction of cell types in DNAm-based CMD.
• Introduced quantitative measures for assessing the uncertainty inherent in CMD results.
• Demonstrated the utility of the developed frameworks for enhancing CMD analysis.

Conclusions:

• The proposed analytical frameworks provide essential tools for defining detection limits in DNAm-based CMD.
• Quantifying uncertainty is critical for robust interpretation of CMD results in EWAS.
• This work advances the reliability and reproducibility of epigenome-wide association studies involving cell mixture deconvolution.