Unveiling Predictive Neural Signatures of Cognitive Adaptability in Aging Bats: A Multi-Region DTI and Machine Learning Approach

Abstract: To understand how brain structure predicts cognitive adaptability in aging, moving beyond simple decline, we investigated predictive neural signatures in Egyptian fruit bats. We developed novel Cognitive Adaptability Indices (CAI) from a spatial re-learning task, which revealed a cognitive trade-off where higher scores reflected better long-term memory but poorer short-term flexibility. For 31 bats, we extracted Mean Diffusivity (MD) from 82 brain regions using Diffusion Tensor Imaging, integrating this with epigenetic age, sex, and origin colony. A machine learning framework, employing ElasticNet and Random Forest regression with Leave-One-Out Cross-Validation, was used to predict CAI. While static features poorly predicted CAI (negative cross-validated R-squared), indicating substantial individual variability in cognitive strategy, we uncovered significant age-modulated brain-behavior relationships. Specifically, ElasticNet regression identified negative interaction effects between epigenetic age and MD in brain regions 9, 22, and 23. This indicates that in older bats, reduced microstructural integrity in these regions is more strongly associated with a cognitive strategy favoring short-term adaptability. Our findings highlight a dynamic reshaping of brain-behavior relationships across the lifespan, where age-related changes in specific neural substrates influence an individual's cognitive strategy rather than simply causing uniform decline. \

Subjects:	q-bio.NC; q-bio.QM
Cite as:	PX:2508.00043