GROWell WGBS Comethylation Analysis

Bioinformatics analysis for identifying co-methylation networks associated with maternal metabolic outcomes in the GROWell clinical trial, focusing on postpartum weight retention (PPWR) and adverse pregnancy outcomes (APOs).

Project Overview

This project applies a co-methylation network framework (comethyl) to whole-genome bisulfite sequencing (WGBS) data derived from participant-collected dried blood spots (DBS) in the GROWell (Goals for Reaching Optimal Wellness) study.

The goal is to:

Identify modules of correlated DNA methylation regions Associate these modules with: Metabolic outcomes (PPWR) Clinical outcomes (APOs) Behavioral and environmental covariates

Data Description

Data type(s): Whole-Genome Bisulfite Sequencing (WGBS)

Study/cohort: GROWell (Goals for Reaching Optimal Wellness)

Sample type: Dried Blood Spots (DBS)

Timepoints:

1. Baseline (early pregnancy)
2. 36–38 weeks gestation
3. ~3 months postpartum

Genome build: hg38

Quick start

Clone

git clone https://github.com/dreusebio/wgbs_growell_comethylation_analysis.git
cd wgbs_growell_comethylation_analysis

---

Repository structure

• data/ - raw + processed data + codebooks, data dictionaries, provenance notes (not commited)
• scripts/ — entrypoints + SLURM submit scripts
• analysis/ — downstream statistics/figures + configuration + sample sheets + generated logs
• docs/ — methods + workflow documentation
• results/ — generated outputs
• test/ — Test data and vignettes

---

#Pipeline Overview (comethyl)

The workflow follows a structured, reproducible pipeline:

1. Preprocessing Input: Bismark cytosine reports (CpG_report.txt.gz) Coverage filtering CpG clustering (≥3 CpGs within ≤150 bp)
2. Region Filtering Minimum coverage threshold Variability filtering (e.g., SD > 0.05–0.08)
3. Methylation Matrix Construction Region-level methylation aggregation Output: Region_Methylation.rds
4. Adjustment for Confounders

Three analysis strategies:

v1_all_pcs → adjust for all PCs v2_exclude_outcome_exposure_pcs → remove biologically relevant PCs v3_technical_pcs_only → adjust only technical variation 5. Network Construction Soft-threshold selection (getSoftPower) Adjacency → TOM → clustering Module detection (WGCNA-based) 6. Module–Trait Association Correlation: bicor (robust, preferred) pearson (optional) Outputs: Heatmaps Dot plots Scatter plots 7. Functional Annotation GREAT (preferred) Offline annotation (EnsDb / TxDb fallback) Enrichment: GO KEGG ClinVar / GWAS (optional)

---

Reproducibility standard

See docs/lab_reproducibility_standard.md.

---

License

MIT License (see LICENSE)