Related Experiment Video
Updated: Aug 31, 2025
10:44
In Vitro Selection of Engineered Transcriptional Repressors for Targeted Epigenetic Silencing
Published on: May 5, 2023
1.5K
Novel feature selection methods for construction of accurate epigenetic clocks
Adam Li1, Amber Mueller1, Brad English1
1Blavatnik Institute, Dept. of Genetics, Paul F. Glenn Center for Biology of Aging Research at Harvard Medical School, Boston, Massachusetts, United States of America.
Plos Computational Biology
|August 19, 2022
Summary
Novel feature selection methods create accurate epigenetic clocks for predicting biological age and health outcomes. These new methods outperform existing models and use fewer CpG sites, offering valuable tools for longevity research.
Area of Science:
- Biogerontology
- Epigenetics
- Bioinformatics
Background:
- Epigenetic clocks accurately predict age and health by analyzing DNA methylation patterns.
- Current methods for constructing epigenetic clocks often use all CpG sites, lacking optimization.
- Efficiently building accurate epigenetic clocks is crucial for longevity research.
Purpose of the Study:
- To develop novel feature selection methods for constructing more efficient and accurate epigenetic clocks.
- To identify a minimal set of CpG sites for reliable age prediction.
- To compare the performance of newly developed clocks against existing models.
Main Methods:
- Applied advanced feature selection techniques, including neural networks and genetic algorithms, to human whole blood methylation data (~470,000 CpGs).
- Developed and validated epigenetic clocks using selected CpG sites.
- Compared performance against established epigenetic clocks (Hannum, Horvath, Weidner) on external datasets.
Main Results:
- Developed clocks predicting age with R2 scores > 0.73.
- Achieved a high R2 score of 0.87 using only 35 CpG sites.
- A clock using the five most frequent CpG sites achieved an R2 score of 0.83.
- New clocks demonstrated superior predictive accuracy compared to Hannum, Horvath, and Weidner clocks on validation datasets.
- Identified potential gene regulatory regions associated with selected CpGs.
Conclusions:
- Novel feature selection algorithms enable the creation of accurate and generalizable epigenetic clocks with a reduced number of CpG sites.
- These optimized clocks serve as powerful tools for assessing biological age and the efficacy of longevity interventions.
- The findings provide a foundation for future aging studies targeting identified gene regulatory regions.
