LFS-age-of-onset

Main folders

1) Scripts

This folder contains the main sub folders for the analysis.

1A) predict_age

This is the main folder in the repository and holds all the scripts used in the main pipeline.

a) Strategy = With Age removal - First principle component removed and data reconstructed.

Run in this order:

First script: read_in_450k.R - this script reads in the .IDAT files and clean the data. it creates 4 main datasets.

1. cases_450_beta.rda (can be m_value instead of beta). These are LFS cancer patients that had methylation generated from 450k array.
2. cases_wt_450_beta.rda - These are cancer patients that don't have LFS (family members. Also cases means cancer patients and controls are cancer free. the wt stands for wild type).
3. controls_450_beta.rda - These are LFS patients with no cancer yet.
4. controls_wt_450_beta.rda - Theses patients that don't have cancer or LFS (family members).

Second Script: read_in_850k. - The same exact thing is done for this script, it creates

1. cases_850_beta.rda
2. cases_wt_850_beta.rda
3. controls_850_beta.rda
4. controls_wt_850_beta.rda

• The only difference is this .rda files with '850' come from IDAT files generated on an 850k array, not a 450k array.

Third Script: get_g_ranges.R

• This script simply reads in generic methylation data to grab genomic locations. We need those locations for the next script to run bumphunter.

Fourth Script: get_pca.R

• This script reads in all they data and uses a function called get_pca to view the batch effects from technology - 450k vs 850k. This script also runs bumphunter to remove the batch effects. It saves a combat version of each dataset and also a version that doen't remove the batch effects (to compare later with the others). I experimented with a lot of combat strategies It also combines the data an

Fifth Script: prepare_data.R

• This script reads in all the data generated at this point: normal data prepare's the data for modeling, buy removing duplicates and running bumphunter to take the subset of probes that are differentially methylated between con_wt (dataset of non cancer non LFS patients) and con_mut (data of non cancer LFS patients). This subset is 'LFS' specific and cuts down the dimentionality of the data significantly).

2) Data

This folder is in the .gitignore file so it's not on github.