Brain aging is a multifaceted process that remains poorly understood. Despite significant advances in technology, progress toward identifying reliable risk factors for suboptimal brain health requires realistically complex analytic methods to explain relationships between genetics, biology, and environment. Here we show the utility of a novel unsupervised machine learning technique - Correlation Explanation (CorEx) - to discover how individual measures from structural brain imaging, genetics, plasma, and CSF markers can jointly provide information on risk for Alzheimer's disease (AD). We examined 829 participants (M age: 75.3 ± 6.9 years; 350 women and 479 men) from the Alzheimer's Disease Neuroimaging Initiative database to identify multivariate predictors of cognitive decline and brain atrophy over a 1-year period. Our sample included 231 cognitively normal individuals, 397 with mild cognitive impairment (MCI), and 201 with AD as their baseline diagnosis. Analyses revealed latent factors based on data-driven combinations of plasma markers and brain metrics, that were aligned with established biological pathways in AD. These factors were able to improve disease prediction along the trajectory from normal cognition and MCI to AD, with an area under the receiver operating curve of up to 99%, and prediction accuracy of up to 89.9% on independent "held out" testing data. Further, the most important latent factors that predicted AD consisted of a novel set of variables that are essential for cardiovascular, immune, and bioenergetic functions. Collectively, these results demonstrate the strength of unsupervised network measures in the detection and prediction of AD.

A recent genome-wide association meta-analysis showed a suggestive association between alcohol intake in humans and a common single nucleotide polymorphism in the ras-specific guanine nucleotide releasing factor 2 gene. Here, we tested whether this variant - associated with lower alcohol consumption - showed associations with brain structure and longitudinal ventricular expansion over time, across two independent elderly cohorts, totaling 1,032 subjects. We first examined a large sample of 738 elderly participants with neuroimaging and genetic data from the Alzheimer's Disease Neuroimaging Initiative (ADNI1). Then, we assessed the generalizability of the findings by testing this polymorphism in a replication sample of 294 elderly subjects from a continuation of the first ADNI project (ADNI2) to minimize the risk of reporting false positive results. The minor allele - previously linked with lower alcohol intake - was associated with larger volumes in various cortical regions, notably the medial prefrontal cortex and cingulate gyrus in both cohorts. Intriguingly, the same allele also predicted faster ventricular expansion rates in the ADNI1 cohort at 1- and 2-year follow up. Despite a lack of alcohol consumption data in this study cohort, these findings, combined with earlier functional imaging investigations of the same gene, suggest the existence of reciprocal interactions between genes, brain, and drinking behavior.

Alzheimer's disease (AD) involves a gradual breakdown of brain connectivity, and network analyses offer a promising new approach to track and understand disease progression. Even so, our ability to detect degenerative changes in brain networks depends on the methods used. Here we compared several tractography and feature extraction methods to see which ones gave best diagnostic classification for 202 people with AD, mild cognitive impairment or normal cognition, scanned with 41-gradient diffusion-weighted magnetic resonance imaging as part of the Alzheimer's Disease Neuroimaging Initiative (ADNI) project. We computed brain networks based on whole brain tractography with nine different methods - four of them tensor-based deterministic (FACT, RK2, SL, and TL), two orientation distribution function (ODF)-based deterministic (FACT, RK2), two ODF-based probabilistic approaches (Hough and PICo), and one "ball-and-stick" approach (Probtrackx). Brain networks derived from different tractography algorithms did not differ in terms of classification performance on ADNI, but performing principal components analysis on networks helped classification in some cases. Small differences may still be detectable in a truly vast cohort, but these experiments help assess the relative advantages of different tractography algorithms, and different post-processing choices, when used for classification.