DNA methylation-based age prediction using cell separation algorithm.

Area of Science:

• Biotechnology
• Bioinformatics
• Computational Biology

Background:

• Predicting biological age from DNA methylation patterns is a key area in aging research.
• Existing methods for age prediction from DNA methylation data often face challenges in accuracy and optimization.

Purpose of the Study:

• To develop and evaluate a novel age prediction method using DNA methylation data.
• To optimize artificial neural network (ANN) models with a new Cell Separation Algorithm (CSA).

Main Methods:

• Developed the Cell Separation Algorithm (CSA), inspired by differential centrifugation, to optimize ANN models.
• Tested the CSA on 25 benchmark functions and applied it to age prediction from blood and saliva DNA methylation data.
• Compared CSA performance against Stochastic Gradient Descent (SGD), ADAM, and Genetic Algorithm (GA).

Main Results:

• The CSA demonstrated significantly better performance than previous methods not utilizing ANN training.
• CSA outperformed SGD and ADAM when applied to ANNs without meta-heuristic optimization.
• CSA achieved comparable or superior results to the Genetic Algorithm (GA) in ANN optimization for age prediction.

Conclusions:

• The Cell Separation Algorithm (CSA) offers a highly effective approach for optimizing ANNs in DNA methylation-based age prediction.
• CSA shows promise for accurate biological age estimation across different sample types, including blood and saliva.
• This novel optimization technique advances the field of epigenetic age prediction.