A superLearner-based pipeline for the development of DNA methylation-derived predictors of phenotypic traits
Data files
Jan 18, 2025 version files 27.66 GB
-
00_SLPCA_Predictor_Guide.html
1.16 MB
-
Childhood_Clock_Inputs.RData
10.69 GB
-
Hannum_Clock_Inputs.RData
10.01 GB
-
PBB_Inputs.RData
6.96 GB
-
README.md
1.88 KB
-
SL_PCA_Clock_Functions.R
18.31 KB
Abstract
Background: DNA methylation (DNAm) provides a window to characterize the impacts of environmental exposures and the biological aging process. Epigenetic clocks are often trained on DNAm using penalized regression of CpG sites, but recent evidence suggests potential benefits of training predictors on principal components.
Results: We developed a pipeline to simultaneously train three epigenetic predictors; a traditional CpG Clock, a PCA Clock, and a SuperLearner PCA Clock (SL PCA). We gathered publicly available DNAm datasets to generate i) a novel childhood epigenetic clock, ii) a reconstructed Hannum adult blood clock, and iii) as a proof of concept, a predictor of polybrominated biphenyl exposure using the three developmental methodologies. We used correlation coefficients and median absolute error to assess fit between predicted and observed measures, as well as agreement between duplicates. The SL PCA clocks improved fit with observed phenotypes relative to the PCA clocks or CpG clocks across several datasets. We found evidence for higher agreement between duplicate samples run on alternate DNAm arrays when using SL PCA clocks relative to traditional methods. Analyses examining associations between relevant exposures and epigenetic age acceleration (EAA) produced more precise effect estimates when using predictions derived from SL PCA clocks.
Conclusions: We introduce a novel method for the development of DNAm-based predictors that combines the improved reliability conferred by training on principal components with advanced ensemble-based machine learning. Coupling SuperLearner with PCA in the predictor development process may be especially relevant for studies with longitudinal designs utilizing multiple array types, as well as for the development of predictors of more complex phenotypic traits.
README: A SuperLearner-based Pipeline for the Development of DNA Methylation-derived Predictors of Phenotypic Traits
https://doi.org/10.5061/dryad.p8cz8w9z3
Description of the data and file structure
DNA methylation-based predictors have emerged as key biomarkers for environmental health research, and it is vital that these predictors maximize the signal-to-noise ratio and are capable of modeling complex exposure-response relationships. Leveraging these methodological advances, we developed a pipeline capable of simultaneously training DNA methylation-based predictors using 3 methods; 1) the traditional approach based on elastic net regression of the CpG matrix, 2) elastic net regression of the principal component matrix, and 3) the ensemble prediction derived from running a SuperLearner model on the principal component matrix.
To compare these 3 clock development methodologies, we trained 3 epigenetic predictors using each of the three training methodologies; 1) a novel childhood clock, 2) a revisiting of the traditional Hannum clock, and 3) a novel predictor of polybrominated biphenyl exposure.
The training model data for each of the predictors is stored in the corresponding RData objects:
- "Childhood_Clock_Inputs.RData"
- "Hannum_Clock_Inputs.RData"
- "PBB_Inputs.RData"
The code necessary for generating predictions from these predictors as well as code for generating new predictors using the methodology developed in this analysis are stored "SL_PCA_Clock_Functions.R
", and a detailed user guide is provided in "00_SLPCA_Predictor_Guide.HTML"
Sharing/Access information
- DNA methylation and phenotypic data used to construct all predictors, as well as most testing datasets, is publicly available on the Gene Expression Omnibus, with the GEO accession numbers specified in the main manuscript.
Methods
Three epigenetic predictors generated using public DNA methylation datasets, including a novel childhood epigenetic clock, a revisiting of the Hannum adult epigenetic clock, and a novel predictor of polybrominated biphenyl exposure.