README.md - EpiGenGO Dissertation Repository

EpiGenGO: Saliva-Based Epigenetic & Genetic Modelling for Personalised Health Insights

*Author: Douglas Acheampong *Course: MSc Data Science *Supervisor:Farukh Aslan *Institution: Coventry University *Year: 2025

Project Overview

EpiGenGO is a research and prototype system designed to explore whether saliva-derived epigenetic (CpG methylation), biometric, and synthetic SNP features can jointly support personalised biological age modelling. The system uses the dataset GSE280465 as the foundational source of CpG data and metadata.

Three structured models were developed:

Model A — Biometric Model

A simple baseline regression model using age + sex only.

Model B — Epigenetic + Genetic Model

A richer feature set combining:

CpG methylation sites extracted from GSE280465
Derived epigenetic age & acceleration
Synthetic SNP variants (FTO, FKBP5, MAOA)
Gene % reactivity logic (prototype architecture)

Model C — Integrated Multi-Layer Model

A combined feature model merging biometrics + epigenetic signals + SNP logic, used to generate biological age predictions and exploratory personalised insights for individuals 0–3.

Repository Structure

EpiGenGO-Dissertation/
│
├── Notebooks/
│   └── data_analysis.ipynb     # Main notebook: preprocessing, models A/B/C, logic
│
├── modelC_output.xls           # Final predictions for individuals 0–3
├── modelC_enriched.xls         # Integrated dataset feeding Model C
├── df_small.xls                # Reduced CpG dataset used for reproducibility
├── meta_buccal.xls             # Cleaned biometric metadata (buccal-only)
├── parsed_metadata_with_tissue.xls
│
├── Images/
│   └── flowcharts, PCA diagrams, logic diagrams
│
└── README.md                   # Project documentation

Note: The full raw GSE280465 dataset (1.1GB) is not included on GitHub due to file size limits. It is available on NCBI GEO: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE280465

Methodology Summary

1. Data Acquisition

Raw methylation matrices and sample metadata were obtained from GEO.
Data was filtered for buccal epithelial cell samples only, using tissue labels.

2. Preprocessing

CpG matrix reduced to a manageable version (df_small.xls) for reproducibility.
Metadata cleaned and aligned by sample ID.
Synthetic SNP variants (FTO, FKBP5, MAOA) were generated using literature-based logic.
Gene reactivity % was calculated using min–max normalisation.

3. Model Development

Model A

Simple linear regression predicting chronological age from biometrics.

Model B

Epigenetic + SNP feature exploration:

Extracted CpG sites
Epigenetic age estimation
Accelerated ageing calculation
SNP–CpG interaction prototype logic

Model C

A unified model integrating all feature layers into a regression architecture. Outputs include:

Predicted biological age
Feature importance (coefficients)
Individualised SNP–epigenetic interaction patterns

How to Run the Notebook

Requirements

Install dependencies:

pip install pandas numpy scikit-learn matplotlib seaborn

Running the Analysis

Open the notebook:

jupyter notebook Notebooks/data_analysis.ipynb

Ensure the small data files (*.xls) are in the project root.
Run all cells sequentially to reproduce Models A–C.
Outputs will include:
- Cleaned datasets
- Model training logs
- Predictions for individuals 0–3
- Data visualisations
- SNP/CpG interplay diagrams

Prototype: SNP–CpG Interaction Logic

This repository contains the logic architecture for:

Synthetic SNP assignment per individual
CpG-driven ON/OFF modulation
Health insight pathways
Gene reactivity normalisation
Integration into the multi-layer Model C

This forms the conceptual backbone of the EpiGenGO mobile health prototype.

Agentic AI Layer (Planned Extension)

A conversational layer will be added later to allow supervisors and examiners to speak directly to the model:

Ask personalised health-style questions
Receive insights for individuals 0–3 (“Eli-0”, “Mara-1”, “Tariq-2”, “Yuna-3”)
Access narrative explanation of SNP/CpG interactions
Prototype clinical-style dialogue powered by rules + embeddings

This extension will be released as: /agentic_layer/epi_chat.py (coming soon)

**Outputs Included

Model C predictions Integrated enriched dataset Metadata tables Flowcharts and conceptual diagrams These files support your dissertation submission and allow examiners to verify reproducibility.

Citation

If using this repository, please cite:

Acheampong, D. (2025). EpiGenGO: A saliva-based epigenetic & SNP modelling prototype for personalised biological ageing. Coventry University.

Contact

For questions or collaboration:

DougScience Email: douglasanim@gmail.com GitHub: https://github.com/DougScience

acheampo7/EpiGenGO_Project

About

Resources

Stars

Watchers

Forks

Releases

Languages