Code for the paper: Machine learning-based DNA methylation score for fetal exposure to maternal smoking: development and validation in samples collected from adolescents and adults
Authors: Sebastian Rauschert, Phillip E Melton, Anni Heiskala, Ville Karhunen, Graham Burdge, Jeffrey M Craig, Keith M Godfrey, Karen Lillycrop, Trevor A Mori, Lawrence J Beilin, Wendy H Oddy, Craig Pennell, Marjo-Riitta Järvelin, Sylvain Sebert, Rae-Chi Huang (in Press DOI 10.1289/EHP6076)
This repository contains the code that was used to train the following four models for the development of an in utero smoke exposure score:
- Random Forest
- Gradient Boosting Machine
- Support Vector Machine
- Elastic Net Regression
The folder code contains the training and testing code, whereas the folder models contains the final gradient boosting machine, random forest and support vector machine caret objects to predict on
new, unseen data.
Further, the code folder contains code to create the Reese et al. score and the Richmond et al. score, which were the "gold standard" models we tested against in our
study.
Further instructions on how to use this repository can be found in the manuscript and in Supplement 2 of the manuscript