Accurate quanti?cation of aging for different organ systems enables detection of any deviation from typical aging and identi?cation of early onset of a disease characterized by accelerated aging, and thus provides opportunities for early intervention. The brain-Predicted Age Difference (brainPAD) is a new clinical informatics framework de?ned as the difference between the brain-derived age and the chronological age of the individual. It has been suggested that brainPAD correlates with physical ?tness, cognitive performance, mild cognitive impairment and Alzheimer's disease (AD). However, several limitations remain in current brainPAD models. Models with large age range yield prediction error larger than 5 years and are often derived from a single-modal imaging feature set (e.g. only structural data). Moreover, due to known in?uence of sex differences on brain morphology, existing models are trained on male and female separately, which may pose challenges in studies with small sample size and in interpretation across models. Finally, because of the phenomenon of regression to the mean, predicted age is overestimated in younger individuals and underestimated in older, which can lead to false positive associations of brainPAD with variables of interest, such as disease status. To address these limitations, this proposal aims to develop and apply novel machine learning algorithms and biomedical software to increase the model accuracy and robustness in three Speci?c Aims: 1) Develop and evaluate a novel feature selection method to identify brain features in?uencing brainPAD; 2) Apply the novel machine learning framework to explore different data types and feature types provided by the Alzheimer's Disease Neuroimaging Initiative (ADNI); 3) Develop and validate integrative methods to create new measures of clinically abnormal aging and compute these measures on the ADNI longitudinal data. Addressing the existing bias and accounting for confounding effects, the new and more accurate measures of a person's age from their brain and epigenetic signatures will provide insights into what causes atypical aging and help predict the onset and individual trajectory of progression in speci?c neurodegenerative diseases such as AD.
The ability to objectively quantify the degree of aging for different organ systems is central to identifying underly- ing biological mechanisms of many neurodegenerative diseases including Alzheimer's. The proposed research provides important methodological improvement to accurately predict accelerated aging and open the doors for early detection of Alzheimer's disease progression.
