Abstract
Recent advances in machine learning have applied novel tools to aging research, yet the relationship between the gut microbiome and epigenetic aging remains underexplored. This proof-of-concept study investigates whether gut microbial composition is associated with biological aging pace independent of chronological age.
Using paired 16S rRNA gene sequencing and DNA methylation data from 123 monocyte-enriched samples in a cohort including Native Hawaiian and Pacific Islander participants, we developed “EpiBiome” models to predict epigenetic age acceleration residuals and DunedinPACE, a DNA methylation biomarker that estimates the instantaneous pace of biological aging. Models predicting residuals of traditional clocks (Horvath, Levine, GrimAge2) showed no predictive signal at either taxonomic rank.
By contrast, the EpiBiome-Accel model for DunedinPACE reached statistical significance at both the species level (R2 = 0.152, Spearman ρ = 0.408, p = 0.012; permutation p < 0.001) and the genus level (R2 = 0.099, permutation p = 0.036). Adding chronological age as a feature did not improve performance (ΔR2 = − 0.046 at species level), indicating age-independence.
SHAP analysis of the species-level ElasticNet model identified Bifidobacterium adolescentis as the dominant contributor and the strongest predictor of decelerated aging, with Succinivibrio dextrinosolvens showing the strongest association with accelerated aging.
These findings reveal specific gut taxa as hypothesis-generating candidates for mechanistic follow-up, rather than as individual-level diagnostic markers.
