RESEARCHERS have shown that AI can identify high-risk individuals using everyday medical and dietary information, with findings also pointing to gut health as a possible driver.
Why Alzheimer’s Risk Prediction Needs New Tools
Early detection of Alzheimer’s disease often depends on costly scans or invasive biomarkers, limiting broad access. Researchers therefore tested whether Alzheimer’s risk prediction could be achieved using questionnaire-based data alone. They also explored whether gut microbiome patterns might help explain links between diet, health history, and later neurodegeneration.
Machine Learning Framework and Study Design
This multi-modal machine learning study analysed 9,832 participants with 120 metadata features spanning five categories: demographic, dietary, lifestyle, nutritional, and medical factors. Four algorithms were trained using 10-fold cross-validation with synthetic minority oversampling technique and externally validated on 1,967 samples. Concurrent 16S rRNA microbiome sequencing data from 2,000 participants enabled exploratory microbial composition analysis. Primary outcomes were predictive accuracy across factor domains, measured by area under the curve. Secondary outcomes included feature importance through Shapley additive explanation analysis and microbiome signatures linked to Alzheimer’s risk prediction.
Results Highlight the Strongest Predictors of Risk
Medical history achieved area under the curve (AUC) of 0.871 and dietary patterns AUC of 0.874, outperforming demographic factors (0.795), lifestyle factors (0.660), and nutritional intake alone (0.569), with p < 0.001 for comparisons. These results indicate Alzheimer’s risk prediction was strongest using medical and dietary data rather than isolated nutrient measures. The model identified influential features including vascular conditions, depression, and eating behaviours. Exploratory microbiome analysis showed dysbiosis markers, including a Prevotella/Bacteroides ratio of 1.921, suggesting potential gut-brain inflammatory pathways relevant to disease development.
Clinical Meaning and Future Prevention Strategies
The findings suggest Alzheimer’s risk prediction may be feasible at community level using low-cost, non-invasive questionnaires rather than specialist testing. This could help identify people for earlier monitoring or preventive interventions focused on modifiable risks. Larger longitudinal studies are still needed to validate causality, refine accuracy, and determine whether microbiome-targeted or dietary strategies can reduce future Alzheimer’s incidence.
Reference
Jabeen T et al. Multi-modal machine learning and gut microbiome pathway analysis for Alzheimer’s risk prediction. Alzheimer’s Dement. 2026;18:e70340.
Featured image: rh2010 on Adobe Stock
