Molecular-enriched functional connectivity derived from neurotransmitter transporter systems is a feasible and biologically informative extension to brain-age modeling, but its added predictive value over structural morphometry was modest and depended on harmonization and atlas alignment.
Key Findings
Results
Single-transporter molecular-enriched functional connectivity explained up to 51% of age variance in healthy adults.
Molecular-enriched connectivity maps were derived using Receptor-Enriched Analysis of functional Connectivity by Targets (REACT) with receptor-density templates for dopamine (DAT), norepinephrine (NET), and serotonin (SERT) transporter systems.
The most predictive transporter varied by dataset, with DAT dominating in the harmonized and common-parcellation settings.
The study analyzed MRI data from 2120 healthy adults (1243 female/877 male; ages 18–90 years) from three public datasets.
Support vector regression models were applied to predict chronological age from molecular-enriched FC features.
Results
Combining all three molecular-enriched transporter maps consistently improved age prediction over any single transporter map, explaining up to 64% of age variance.
The combined three-transporter model (DAT, NET, SERT) outperformed each individual transporter model across datasets.
Explained variance increased from up to 51% (single transporter) to up to 64% (combined transporter maps).
This finding held consistently across the datasets examined.
Results
Structural morphometry remained the strongest single modality for brain-age prediction overall.
Structural MRI features outperformed molecular-enriched functional connectivity as a standalone modality.
In the merged multi-site cohort using a common parcellation, structural morphometry achieved a mean absolute error (MAE) of 6.02 years.
This finding was consistent across the different analytical settings examined in the study.
Results
Adding transporter-enriched functional connectivity to structural morphometry features yielded a small but statistically reliable reduction in prediction error in the harmonized, common-parcellation setting.
In the merged multi-site cohort using a common parcellation, combining transporter-enriched FC with structural features reduced MAE from 6.02 to 5.81 years.
Residual-level paired comparisons across repeated cross-validation confirmed that this improvement is statistically reliable but modest in magnitude.
The improvement was described as supporting 'limited complementarity between the two modalities.'
Results
Parcellation mismatch between modalities obscured the functional connectivity contributions to brain-age prediction.
When different parcellations were applied to structural and functional data, incorporating molecular-enriched FC into brain-age prediction resulted in a 2% higher MAE compared to structural morphometry alone.
In contrast, using a common parcellation for both modalities revealed the complementary benefit of adding molecular-enriched FC.
A common-parcellation analysis was specifically conducted to assess the impact of differing parcellations between modalities.
Methods
Multi-site variability was addressed using ComBat harmonization, with and without Empirical Bayes pooling, and harmonization settings affected which transporter was most predictive.
Data came from three public datasets, introducing multi-site variability that required harmonization.
ComBat harmonization was applied with and without Empirical Bayes pooling.
The most predictive transporter varied by dataset, with DAT dominating specifically in the harmonized and common-parcellation settings, indicating harmonization influenced model outcomes.
What This Means
This research suggests that brain aging can be estimated from MRI scans, and that adding information about neurotransmitter systems — the brain's chemical messenger networks — to standard brain structure measurements provides some additional biological insight. The researchers used a technique called REACT to link resting-state brain activity patterns to the locations of specific chemical transporters for dopamine, norepinephrine, and serotonin, creating what they call 'molecular-enriched' brain maps. They tested whether these maps could predict a person's age in a sample of 2,120 healthy adults ranging from 18 to 90 years old. They found that these molecular-enriched maps alone could explain up to 64% of age-related variation in the brain when all three transporter systems were combined, and that adding them to structural brain measurements modestly but reliably reduced prediction error.
However, the practical gains were limited. Traditional structural MRI — which measures things like brain size and thickness — remained the most powerful single predictor of brain age on its own. Combining the molecular-enriched functional data with structural data only reduced the average prediction error by about 0.2 years (from 6.02 to 5.81 years). Importantly, the researchers found that methodological choices mattered a great deal: when different brain region maps (parcellations) were used for the two types of data, adding the functional data actually made predictions slightly worse, and the best-performing neurotransmitter system differed depending on how the data were processed.
This research suggests that incorporating neurotransmitter system information into brain-age models is scientifically meaningful because it ties predictions to known biological systems, potentially making the models more interpretable. However, it also highlights that careful methodological alignment — including consistent brain region maps and proper handling of data from multiple scanning sites — is essential for realizing any benefit. The findings may be relevant to future studies aiming to understand how specific neurochemical systems contribute to healthy and unhealthy brain aging.
Pinamonti M, Moretto M, Sammassimo V, Castellaro M, Veronese M. (2026). Investigating the Contribution of Molecular-Enriched Functional Connectivity to Brain-Age Analysis.. Human brain mapping. https://doi.org/10.1002/hbm.70627