DNA Methylation Predicts Early Onset of Primary Tumor in Patients with Li-Fraumeni Syndrome

Project Overview

Li-Fraumeni syndrome (LFS) is an autosomal dominant cancer predisposition syndrome. Approximately 80% of LFS patients harbor a germline TP53 mutation, rendering them susceptible to a wide spectrum of early onset malignancies. A comprehensive surveillance regimen termed the 'Toronto Protocol' has recently been adopted for early tumor detection, demonstrating significant improvement in survival among these patients. However, the protocol's "one-size-fits-all" approach fails to consider an individual patient's risk of cancer.

This project has developed a machine learning model that predicts early onset of primary tumors in LFS patients by estimating the probability of cancer onset before the age of six, leveraging a patient's peripheral blood leukocyte methylation profile.

Repository Structure

Preprocessing

/preprocessing

Contains scripts for preprocessing DNA methylation data, including optimal preprocessing and normalization methods, outlier detection, and confounder removal.

Model Development

/model

Contains scripts for feature selection, cross-validation, and model selection. Includes utilities for running models with different seeds and parameters, and combining results.

Analyses

/analyses

Scripts for analyzing model results, including SHAP value analysis, covariate comparisons, age-related probe analysis, model comparisons, and family-based analyses.

Sampling Analysis

/sampling

Contains scripts related to resampling techniques used to address class imbalance and validate model robustness.

Immune Cell Deconvolution

/idol

Scripts for estimating immune cell proportions from methylation data using the IDOL (Identifying Optimal Libraries) method.

Figures

/figures

Contains generated figures and visualization scripts for the project.

Checkpoints

/checkpoint

Model checkpoints and saved model states.

Data Directory

/data

This project uses a centralized data directory structure with the following organization:

• /data/rds/ - For RDS data files
• /data/Output/ - For output files
• /data/Plots/ - For plot files
• /data/Resources/ - For resource files like probe lists
• /data/Data/LFS_450/ - For 450k array data
• /data/Data/LFS_850/ - For 850k array data

Setting up the Data Directory

The project has been updated to use relative paths based on a centralized data directory. Run the setup script to create the necessary directory structure:

./scripts/setup_data_dirs.sh

This script will create the required folder structure and provide instructions for migrating existing data.

Usage

Scripts in this repository are primarily R and shell scripts. Main model execution can be performed using:

./model/runModel.sh

For parallel execution:

./model/runAllModelinParallel.sh

Citation

If you use this code or model in your research, please cite our work (citation details to be added upon publication).