Evaluating transcriptional alterations associated with ageing and developing age prediction models based on the human...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Ageing is associated with molecular changes in DNA methylome and proteome.
• Existing ageing clock models often lack transcriptomic data from human blood.
• Machine learning algorithms can potentially develop accurate age prediction models.

Purpose of the Study:

• To develop and compare age prediction models using human blood transcriptome data.
• To identify genes and biological pathways associated with ageing.
• To evaluate the performance of different machine learning algorithms for age prediction.

Main Methods:

• Utilized blood transcriptome data from 10K Immunomes repository (ages 21-90).
• Applied machine learning algorithms: Least Absolute Shrinkage and Selection Operator (LASSO), Elastic Net (EN), eXtreme Gradient Boosting (XGBoost), and Light Gradient-Boosting Machine (LightGBM).
• Performed differential gene expression analysis, gene ontology, pathway, and disease ontology analysis.

Main Results:

• XGBoost (142 genes) and LightGBM (149 genes) outperformed LASSO (7 genes) and EN (9 genes) in age prediction.
• Gradient Boosting models achieved higher accuracy on training, test, and external validation sets.
• Identified differentially regulated transcripts and associated biological functions related to ageing.

Conclusions:

• Blood transcriptome-based age prediction models offer a feasible method for monitoring biological ageing.
• These models provide molecular insights into the ageing process.
• Further external validation in diverse populations and mechanistic studies are recommended.