Related Experiment Video
Updated: Nov 28, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
7.2K
Autoencoded DNA methylation data to predict breast cancer recurrence: Machine learning models and gene-weight
Laura Macías-García1, María Martínez-Ballesteros2, José María Luna-Romera2
1Department of Citology and Histology, Faculty of Medicine, University of Seville, Seville, Spain.
Artificial Intelligence in Medicine
|November 30, 2020
Summary
This study introduces a novel method using autoencoders to summarize DNA methylation data for breast cancer recurrence prediction. The approach identifies key genes associated with recurrence, aiding in better patient stratification and treatment strategies.
Area of Science:
- Genomics
- Computational Biology
- Oncology
Background:
- Breast cancer is a leading cause of cancer death in women, with recurrence and metastasis posing significant threats.
- DNA methylation data is crucial for understanding breast cancer, but its high dimensionality presents challenges in feature selection.
- Autoencoders (AEs) offer a promising approach for nonlinear feature fusion in high-dimensional data.
Purpose of the Study:
- To develop a procedure for summarizing DNA methylation data using autoencoders.
- To identify a limited set of relevant genes for characterizing breast cancer recurrence.
- To leverage CpG site values for generating new predictive features.
Main Methods:
- Utilized autoencoders (AEs) for nonlinear feature fusion of DNA methylation data.
- Applied survival analysis and a weighted gene ranking based on CpG site distribution.
- Selected a dataset from The Cancer Genome Atlas (TCGA) portal.
- Employed an AE with a single hidden layer for feature generation.
Main Results:
- Generated new features from CpG site values for patients with and without recurrence.
- Identified a set of genes significantly related to breast cancer recurrence.
- Validated the relevance of identified genes through literature and enrichment analysis.
Conclusions:
- The proposed autoencoder-based method effectively summarizes DNA methylation data for breast cancer recurrence.
- The identified genes provide valuable insights into the mechanisms of breast cancer recurrence.
- This approach aids in better characterization and potential prediction of breast cancer recurrence.
