Integrating imaging and omics for enhanced subtyping of mild cognitive impairment associated with Alzheimer's disease

Mild Cognitive Impairment (MCI), considered the prodromal stage of Alzheimer's disease (AD), is a heterogeneous condition characterised by mild but measurable cognitive decline. However, not all individuals with MCI follow the same trajectory-some remain stable, while others progress rapidly to AD. Understanding variation in clinical, molecular, and imaging features is crucial for reducing disease heterogeneity, improving prognosis, and developing targeted interventions. This study aims to increase MCI subtyping resolution by generating enriched individual-level profiles through the integration of imaging and omics data, facilitating precision medicine approaches for AD prevention and treatment.

We used data from the Alzheimer's Disease Neuroimaging Initiative (ADNI), including structural MRI, CSF peptidomics/proteomics, and clinical indices. Using a multi-modal integration and clustering framework, we identified distinct MCI subgroups, characterised by clinical and neuropsychological scores, AD biomarkers, biological pathway enrichments, and imaging patterns. We further employed supervised multi-modal integration and correlation analyses to explore the links between imaging, peptidomic/proteomic and clinical features within each subgroup. Additionally, we labelled individuals by future conversion to AD and analysed longitudinal cognitive function (CDRSB and MMSE scores). Finally, we performed in silico drug repurposing to identify candidate drugs targeting each subgroup's molecular profile.

(1) Multi-modal integration revealed two distinct MCI subgroups. (2) The Resilient Neuronal Hyperplasticity subgroup was characterised by elevated markers of neuronal plasticity, minimal brain atrophy and cortical thinning, better clinical scores, and upregulated peptide/protein markers associated with less severe structural changes. In contrast, the Vulnerable Neurodegenerative subgroup exhibited AD-like disturbances, pronounced atrophy and cortical thinning, primarily affecting executive functions, and downregulation of peptide/protein markers linked to significant structural changes. (3) Future conversion analysis revealed the second subgroup predominantly comprised fast converters, while the first predominantly consisted of stable individuals. (4) Longitudinal cognitive analysis showed a more pronounced decline in the second subgroup compared to the first. (5) Drug repurposing identified both shared and subgroup-specific candidate compounds aligned with the underlying pathologies.

This study delineates two MCI subgroups, using multi-modal integration, offering insights into disease heterogeneity and laying the foundation for precision medicine and AI-driven strategies in MCI and AD research and clinical care.