Increasing temporal sensitivity of omics association studies with epigenome-wide distributed lag models

Area of Science:

• Environmental epigenetics
• High-dimensional data analysis
• Statistical modeling

Background:

• Current methods for identifying temporal effects of exposures in omics data have limitations in high dimensions and may miss time-varying associations.
• Existing approaches often use separate models for screening and temporal window identification, potentially reducing sensitivity.

Purpose of the Study:

• To introduce a novel statistical approach, the epigenome-wide distributed lag model (EWDLM), for efficient screening of timing-specific effects in high-dimensional omics data.
• To combine traditional false discovery rate (FDR) methods with distributed lag models (DLMs) to identify susceptible time windows for exposures.

Main Methods:

• The EWDLM integrates DLM effect estimates with FDR control by marginalizing DLM estimates over time and correcting for multiple comparisons.
• The approach was evaluated through simulations examining the timing-specific effects of air pollution on DNA methylation.
• A real-world analysis assessed the association between prenatal fine particulate matter exposure and DNA methylation at age 12 years.

Main Results:

• Simulations demonstrated that EWDLM achieved increased sensitivity in detecting associations limited to specific exposure time periods compared to traditional two-stage methods.
• The real-world analysis identified 353 cytosine-phosphate-guanine (CpG) sites where DNA methylation at age 12 years was significantly associated with prenatal fine particulate matter exposure.
• EWDLM proved efficient and sensitive in screening for exposure-specific temporal effects in epigenomic data.

Conclusions:

• The epigenome-wide distributed lag model (EWDLM) offers an improved method for identifying temporal windows of effect in high-dimensional omics studies.
• EWDLM enhances the ability to detect associations between environmental exposures and epigenetic modifications localized to specific developmental periods.
• This novel approach facilitates a more sensitive and efficient screening of epigenomic datasets for time-varying exposure effects.