4,068 posters, 219 videos, 101 topics, 69 sessions, 14,365 authors, 6,128 institutions

Alzheimer's Association International Conference 2026

July 12-15, 2026 | London, United Kingdom, and Online

12022

Long Term Validation Of Neuronal Density With A Mathematical Model Of Brain Health

View Poster Contact Presenter

Poster Presenter

HALIMA SADIA

Authors

HALIMA SADIA

Affiliations

Universite Laval

Part of Topic

Basic Science and Pathogenesis -> Molecular and Cell Biology -> Neurodegeneration and neuroprotection

BACKGROUND
Understanding how biological processes drive neurodegeneration is essential for improving the prediction and monitoring of Alzheimer’s disease progression. Our group developed a multiscale mathematical model that simulates the 50-year trajectory of brain health via 19 coupled ordinary differential equations with 75 parameters [1]. The model describes interactions between amyloid-β accretion, tau phosphorylation, inflammation processes, and neuronal degeneration. In previous work [2], we performed a local sensitivity analysis by perturbing all 75 parameters and pick up the 30 parameters that have the greatest influence on three selected outcomes (Ab plaques, tau concentration, and neuronal density). This study focuses on real-world validation of the mathematical model by first optimizing these parameters using the ADNI cohort, then applying the optimized model to predict neuronal density (expressed as estimated neuronal mass in grams) in the UK Biobank cohort.

MODEL DESCRIPTION

Figure 1: Proposed in Chamberland et al. 2024 [1], the model is composed of 19 ordinary differential equations (ODEs) and 75 parameters, covering entities such as the formation of Aβ plaques; tau tangles; neuronal death; and microglia/macrophages interactions in neuroinflammation.

METHODS

Two different approaches are used for this analysis:

1.Global Model Optimization (ADNI)

2. External Validation (UK Biobank

Estimated Brain Volume Calculation

•Model initialized at Age 30 with participants’ biological Sex and initial brain volume, estimated using the Total Intracranial Volume (eTIV) from MRI as follows:

GUIDING QUESTIONS

1.How well do our simulated trajectories predict real-world neuronal density?

2.Can a mechanistic model accurately capture age-related brain atrophy?

RESULTS

•Parameter optimization on ADNI yielded a strong baseline correlation of r = 0.73 (n = 211), which generalized beautifully to the independent UK Biobank cohort (n = 35,899) with an exceptional out-of-the-box correlation of r = 0.85.

•The model higher accuracy in the UK Biobank is attributed to a shorter average simulation timeline (34 years vs. 44 years in ADNI), which naturally minimized long-term ordinary differential equation (ODE) simulation drift.

CONCLUSION

•Our validated multiscale mathematical model successfully links mechanistic biological pathways with biological plausibility when compared to real-world population data.
•Calibrated on ADNI and validated on the UK Biobank, the model precisely predicts real-world macroscopic neuronal density over a 50-year lifespan.
•Exceptional predictive performance across 35,899 unseen individuals demonstrates powerful generalizability and population-level reliability without requiring any extra model tuning.

REFERENCES

[1] Chamberland, E. et al. (2024). Frontiers in Neuroinformatic, 18, 1348113.

[2] Moravveji S., Sadia H., et al. (2025) Frontiers in Neuroinformatics, 19, 2025.

[3] O’Brien LM,. Et al. (2011). Psychiatry Research: Neuroimaging. 2011;193(2):113-22.

Comments

Email is required but is not shown publicly.

Sort comments

Loading comments...