DNA Methylation Biomarkers-Based Human Age Prediction Using Machine Learning.

Area of Science:

• Forensic Science
• Epigenetics
• Biomarker Discovery

Background:

• DNA methylation patterns are closely associated with chronological aging.
• Existing age prediction models often overlook the impact of diseases on epigenetic aging.
• Accurate age estimation from biological evidence is crucial for forensic investigations.

Purpose of the Study:

• To develop and evaluate an age prediction model utilizing DNA methylation biomarkers.
• To create a model applicable to both healthy and diseased biological samples.
• To identify reliable CpG sites correlated with age across diverse health statuses.

Main Methods:

• Utilized publicly available datasets comprising 454 healthy and 400 diseased samples (ages 1-89).
• Identified six CpG sites with high age correlation using Pearson's correlation coefficient.
• Developed age prediction models using Multiple Linear Regression, Support Vector Regression, Gradient Boosting Regression, and Random Forest Regression, with an 80:20 train-test split.

Main Results:

• Random Forest Regression achieved the best performance, followed by Gradient Boosting Regression.
• For healthy samples, the Random Forest model yielded a Mean Absolute Deviation (MAD) of 2.51 years (training) and 4.85 years (testing).
• For diseased samples, the Random Forest model yielded a MAD of 3.83 years (training) and 9.53 years (testing).

Conclusions:

• The proposed DNA methylation-based model effectively predicts age in both healthy and diseased individuals.
• The findings highlight the potential of epigenetic biomarkers for age estimation in forensic contexts.
• The model's performance indicates its utility for forensic casework involving biological evidence.