Repository navigation
Home
Welcome to the Diverse Epigenetic Epidemiology Partnership analysis pipeline. This wiki will guide you through the different stages of the analysis. We recommend that you read this page before running the pipeline to familiarize yourself with its design, contents, and usage, as well as how to use this wiki and what to do if you encounter issues.
A reminder of our participation requirements can be found on this page.
This flowchart provides an overview of the pipeline workflow, encompassing multiple stages within the DEEP scope.
- Blue chunk indicates tasks executed by each DEEP cohort on their respective local HPC servers.
- Orange chunk represents tasks performed by the DEEP core team on the Bristol HPC.
- Green sections show the Stage where scripts are ready to be run.
- Yellow sections show the Stage currently under development.
- White sections show the Stage planned in DEEP.
Our collaborative framework follows the principle of federated analysis, where individual-level data remain on local servers, and only non-disclosive summary statistics are shared across cohorts. Each analyst runs the pipeline for their cohort on their own server, e.g., a High-Performance Computing (HPC) server, and uploads encrypted results (incl. summary statistics, plots) to an SFTP server in Bristol.
The pipeline is designed to make it as automatic and reproducible as possible, minimising coding and input required by users ("analysts"). This makes the pipeline easy to execute, avoids duplication of effort, maintains consistency across cohorts, and reduces potential for errors. Pipeline stages are separated into numbered modules. Modules need to be completed in sequence. This Script Flowchart represents the dependencies between the modules so you can see what order you need to run them in.
Analyst input is minimised by using a config file and pre-formatted genetic, methylation, phenotype, and covariate data. This config file includes all cohort-specific parameters, such as file paths, properties of the data, thresholds, etc. Details on data formatting are provided in Data preparation prior to pipeline Section in the Sidebar: Imputed genetic data, Normalised methylation data, Covariate data, and Phenotype data.
The pipeline begins with a series of sequential modules that verify the correct software requirements, genetic data, DNA methylation data, phenotype data, and covariate data. This generates several outputs that analysis modules can use later in the pipeline.
Perform QC without removing individuals, such as: checking variant and sample missingness rates, sex-mismatch, performing positive/negative control for mQTL and EWAS.
Collect summary-level information from QC stage 1, and perform QC by removing individuals for cross-cohort analysis
Perform a full GWAS on every methylation probe available in the sample. Variations of this analysis will be performed, including:
- a standard mQTL analysis using bestguess imputed SNPs covering the full surface using new software
- cell-type interacting mQTL analyses for abundant cell types
- a var-mQTL analysis, which looks for SNPs that influence the variance of a probe.
We will be conducting EWASs on a range of complex traits:
- Prevalent Type 2 Diabetes
- Body Mass Index
- Blood Pressure
- Blood glucose
- HbA1c
- Blood lipids
- Smoking (y/n)
- Alcohol intake
- Child height
- Child lipids
- Mothers smoking
We will conduct EWASs using the following models:
- standard EWAS
- Variance EWAS
- Cell type interaction EWAS
- DMR analyses
Please note, not all of these modules are ready to be run. Those that are have been included in this repository on the main branch. The others are being developed and will be released once they are ready. You will be notified when new analyses are ready for you to run.
This wiki is structured to guide you through each stage of the analysis. There is one page per module. At the top of each page is a standard header that indicates:
- Module Status: Whether the module is ready to be run.
- Prerequisites: Which scripts must be completed before running this module.
- Next Steps: Instructions for uploading results at the end of the process. Each page describes the analysis performed in the module and provides the necessary commands to implement the scripts.
If you encounter any bugs, errors in the documentation, or technical problems while running the code, please use the Issues section of this repository. When reporting an issue, please include the module name and any error messages to help us resolve the problem more efficiently. If you have developed a fix or an improvement, we welcome your contributions via Pull Requests. Please ensure your PR includes a clear description of the changes made.