Table 1 Challenges and potential solutions for multimodal diversity in precision brain health
From: Computational whole-body-exposome models for global precision brain health
Domain | Challenges and limitations | Potential solutions |
|---|---|---|
1. Data representation | - Underrepresentation of global and minority populations in datasets. - Bias toward US/European ancestry in genetics, imaging, and clinical cohorts. - Reduced generalizability of predictive models due to narrow population samples. | - Expand and integrate datasets to include diverse ancestries, socioeconomic backgrounds, and environmental contexts. - Promote equity-driven data collection strategies, particularly in the Global South. - Implement deep phenotyping across different populations to capture heterogeneity and unique risk/protective factors. |
2. Multimodal integration of extracerebral measures | - Brain health research overly focused on brain-specific measures, neglecting systemic, exposome, and whole-body interactions. - Limited frameworks for modeling biology- environment interactions. - Sparse integration of omics, cardiovascular, metabolic, and environmental data in brain health models. | - Systematic integration of multimodal data, combining brain, body, and exposome metrics. - Leverage organ clocks, multi-omics, and exposome profiling to capture systemic and environmental impacts on brain health. - Adopt ecological and One Health perspectives to link social, environmental, and biological domains in brain health. |
3. Computational and theoretical structures supporting diversity | - Existing models mainly based on low-dimensional, single-domain data. - Poor adaptability to heterogeneity and individual variability. - Lack of temporal integration between brain dynamics, extracerebral influences, and environmental exposures in predictive modeling. | - Develop multimodal metamodels combining: (i) spatiotemporal dimensionality reduction to generate low-dimensional latent representations of brain-body-exposome dynamics; (ii) biophysical whole-brain generative models to integrate multimodal biological measures; (iii) embodied models linking extracerebral (organ, exposome) states to brain dynamics; and (iv) Digital twins and probabilistic frameworks to model heterogeneous individual trajectories across the lifespan. |
