Behavioral Risk Prediction in Acute Mental Health Inpatient Unit

Identifier:  TFM:2526

Handle: https://hdl.handle.net/20.500.11797/TFM2526

Authors:  Barroso Gil, Andrea

Abstract:
This study develops a machine learning framework to predict chronological age from gut microbiome composition in healthy individuals. By analyzing species-level abundance profiles and employing four feature selection strategies, the model identifies aging-related signatures, with permutation feature importance (PFI) proving most effective. The optimal model (CatBoost regressor with PFI) achieved a mean absolute error of 9.17 years, highlighting key microbiome contributors like Coprococcus. Age-related shifts, such as increased Coprococcus comes abundance, suggest more efficient short-chain fatty acid production in older individuals, offering insights into microbiome resilience and potential diagnostic tools for healthy aging.