FeSTwo, a two-step feature selection algorithm based on feature engineering and sampling for the chronological age...

Area of Science:

• Forensic Science
• Bioinformatics
• Genomics

Background:

• Accurate chronological age determination is crucial in forensic science.
• Methylomic features hold potential for improving age prediction models.
• Existing feature selection methods often overlook feature engineering's impact on regression performance.

Purpose of the Study:

• To develop and evaluate novel feature engineering strategies for methylomic data.
• To propose a resampling-based feature selection algorithm, FeSTwo, for enhanced age regression.
• To assess the performance of FeSTwo against existing algorithms using independent datasets.

Main Methods:

• Proposed four feature engineering strategies to transform original methylomic features.
• Developed FeSTwo, a resampling-based feature selection algorithm.
• Evaluated FeSTwo on age regression tasks, comparing it with parallel algorithms.
• Validated age prediction performance on an independent dataset.

Main Results:

• FeSTwo significantly improved the regression performance of the age prediction model.
• The proposed FeSTwo algorithm outperformed previous parallel algorithms.
• FeSTwo-selected features demonstrated robust age prediction on an independent dataset.
• Achieved over 8% reduction in root-mean-square error (RMSE) using only 70 features.

Conclusions:

• FeSTwo offers a superior approach to methylomic feature selection for chronological age determination.
• Feature engineering combined with resampling-based selection enhances age prediction accuracy.
• The developed method provides a more accurate and efficient tool for forensic age estimation.