Back

Residual Partial Least Squares Learning: Brain Cortical Thickness Simultaneously Predicts Eight Non-pairwise-correlated Behavioural and Disease Outcomes in Alzheimer's Disease

Chen, O. Y.; Vu, D. T.; Diaz, C. S.; Bodelet, J. S.; Phan, H.; Allali, G.; Nguyen, V.-D.; Cao, H.; He, X.; Muller, Y.; Zhi, B.; Shou, H.; Zhang, H.; He, W.; Wang, X.; Munafo, M.; Trung, N. L.; Nagels, G.; Ryvlin, P.; Pantaleo, G.

2024-03-13 neuroscience
10.1101/2024.03.11.584383 bioRxiv
Show abstract

Alzheimers Disease (AD) is the leading cause of dementia, affecting brain structure, function, cognition, and behaviour. While previous studies have linked brain regions to univariate outcomes (e.g., disease status), the relationship between brain-wide changes and multiple disease and behavioural outcomes of AD is still not well understood. Here, we propose Residual Partial Least Squares (re-PLS) Learning, an explainable and generalisable framework that models high-dimensional brain features and multivariate outcomes, accounting for confounders. Using re-PLS, we map the many-to-many pathways between cortical thickness and multivariate AD outcomes; identify neural biomarkers that simultaneously predict multiple outcomes; control for confounding variables; conduct longitudinal AD prediction; and perform cross-cohort AD prediction. To evaluate its efficacy, we first carry out within-cohort cross-subject validation using ADNI data, and further examine its reproducibility via between-cohort cross-validation using ADNI and OASIS data. Together, our results unveil brain regions jointly but differentially predictive of distinctive cognitive-behavioural scores in AD.

Matching journals

The top 5 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.