Exploratory Multi-Modal Investigation of Brain Microstructure and Epigenetic Aging in Egyptian Fruit Bats: Identifying Phenotypes of Resilience and Vulnerability

Abstract: Aging trajectories are highly heterogeneous, with some individuals exhibiting remarkable cognitive resilience while others show vulnerability, making the understanding of their multi-modal signatures crucial. This study aimed to identify brain-behavioral correlates of epigenetic aging and stratify distinct aging phenotypes in a cohort of long-lived Egyptian fruit bats (Rousettus aegyptiacus). We initially sought to integrate a novel Diffusion-weighted Signal Variability (DW-SV) metric from 4D MRI with advanced behavioral entropy and efficiency measures to predict DNA methylation (DNAm) age. However, due to the 3D format of the provided MRI data, DW-SV calculation was not possible, leading to the use of regional Mean Signal Intensity. Additionally, the planned behavioral metrics exhibited no variance across subjects and were consequently excluded. For a final cohort of 31 bats, an Elastic Net regression model, utilizing regional Mean Signal Intensity and demographic factors, was trained using Leave-One-Out Cross-Validation to predict DNAm age. The model demonstrated poor predictive performance (R-squared = -0.101, Mean Absolute Error = 1.405 years), indicating that the available neuroimaging features were not strong predictors of epigenetic age in this dataset. Despite this, an exploratory analysis of model coefficients highlighted specific brain regions whose mean signal intensity was weakly associated with epigenetic age. Furthermore, based on the discrepancies between actual and predicted DNAm age, bats were descriptively stratified into 'Resilient' and 'Vulnerable' phenotypes, and their respective neuroimaging profiles were characterized. These findings underscore the challenges in multi-modal data integration for aging research when confronted with data limitations, suggesting that while the current features were insufficient for robust prediction, the developed framework for phenotype identification remains valuable for future studies with richer datasets.

Subjects:	q-bio.NC; q-bio.QM
Cite as:	PX:2508.00041