Network-based prediction of spreading of tau pathology in Alzheimer’s disease: implications for clinical trials

Informations projet

networkAD

N° de convention de subvention: 101153323

DOI

10.3030/101153323

Projet clôturé

Date de signature de la CE 10 Avril 2024

Date de début 1 Mai 2024

Date de fin 30 Avril 2026

Financé au titre de

• Marie Skłodowska-Curie Actions (MSCA)

Coût total

Aucune donnée

Contribution de l’UE

€ 222 727,68

Coordonné par

LUNDS UNIVERSITET
Sweden

Periodic Reporting for period 1 - networkAD (Network-based prediction of spreading of tau pathology in Alzheimer’s disease: implications for clinical trials)

Période du rapport: 2024-05-01 au 2026-04-30

Résumé du contexte et des objectifs généraux du projet

Alzheimer’s disease is the most common cause of dementia and a major health and societal challenge. One of its defining features is the accumulation of tau pathology, which is closely associated with neurodegeneration, cognitive decline and disease progression. However, tau does not accumulate in the same spatial pattern or at the same rate in every person. This heterogeneity complicates prognosis and may reduce the efficiency of clinical trials when participants with different biological trajectories are analysed together.

Tau pathology is thought to spread partly along networks of interconnected brain regions. Most previous studies have used connectivity maps averaged across groups, potentially overlooking meaningful differences in each person’s brain organisation.

networkAD aimed to develop personalised computational neuroimaging methods for studying the spatial distribution and progression of tau pathology. The project investigated whether individual functional-connectivity patterns can improve the characterisation of patient-specific tau-PET topographies and whether baseline connectivity contains information about future tau distribution.

The longer-term pathway to impact is to support more precise disease staging, monitoring and participant characterisation in therapeutic studies. Better identification of biologically comparable patients could ultimately improve clinical-trial design and the assessment of treatment effects.

Travail effectué depuis le début du projet jusqu’à la fin de la période considérée dans le rapport et principaux résultats atteints jusqu’à présent

The main scientific work used resting-state functional MRI and tau-PET data from 805 participants across the Alzheimer’s disease continuum in the Swedish BioFINDER-2 study. Follow-up tau-PET was available for 422 participants.

A personalised hybrid functional-connectivity model was developed by combining stable population-level connectivity with individual-specific connectivity information. The relationship between connectivity and tau-PET was evaluated at ten spatial resolutions, ranging from 100 to 1,000 cortical regions.

The hybrid model explained individual tau-PET patterns better than either population-average connectivity or raw individual connectivity alone. Its added value was particularly clear in symptomatic participants and at finer spatial resolutions. It also captured individual tau topographies better than conventional cohort-level tau patterns in clinically relevant groups.

Analyses using amyloid-PET did not show the same advantage, indicating relative specificity of the connectivity relationship to tau. Longitudinal analyses further showed that baseline hybrid connectivity explained follow-up tau-PET patterns better than population-average connectivity.

The main manuscript is under revision following peer review at Nature Communications. The revision includes spatial null models, cross-validated regularised prediction, coefficient-stability tests and sensitivity analyses using thresholded connectivity. The principal findings have also been replicated in an independent A4 cohort of more than 300 participants. Replication in ADNI cohort (+700 participants) is ongoing.

The manuscript submission included a code repository, which is being improved for public release. The final publication is intended to be accompanied by an open-source GitHub repository. Non-personalised group-level derivatives, including canonical tau maps, have already been shared with external researchers and are intended for wider public release.

A related supervised Master’s project developed a subject-specific multiscale method for quantifying hemispheric asymmetry in cortical maps using geometric eigenmodes. The method was validated using functional MRI and tau-PET data from 545 participants. The work was presented as a poster at IEEE ISBI 2025 and published as a peer-reviewed proceedings paper in IEEE Xplore. The thesis received an institutional award for outstanding interdisciplinary work combining life sciences and informatics.

Progrès au-delà de l’état des connaissances et impact potentiel prévu (y compris l’impact socio-économique et les conséquences sociétales plus larges du projet jusqu’à présent)

The project advanced the state of the art by showing that individual functional-connectivity information adds explanatory value beyond conventional group-average brain networks. The findings indicate that inter-individual variability in connectivity contains relevant information about heterogeneity in tau-PET topography and should not be treated only as measurement noise.

The project also demonstrated that spatial resolution is a major methodological factor. Finer-grained parcellations improved the characterisation of individual tau patterns, particularly in symptomatic participants, suggesting that coarse brain divisions may obscure relevant local relationships between connectivity and pathology.

The hybrid modelling strategy provides a practical way to combine the stability of population-level connectivity with person-specific information. This is particularly important in clinical imaging, where individual functional-connectivity estimates can be noisy.

The longitudinal results show that baseline connectivity is associated with later tau-PET topography. These findings are consistent with network-based accounts of tau progression, although they do not establish a causal propagation mechanism.

External replication in the A4 cohort strengthens the generalisability of the principal finding. Further uptake will require completion of additional external validation, public release of the analysis tools, and future integration with structural connectivity, amyloid burden and regional biological vulnerability. These extensions could support more robust personalised models of disease progression and future clinical evaluation.