Deep learning models for neuroimaging have largely been developed for individual tasks, limiting knowledge transfer across applications. Here we introduce GenFAR, a modular deep learning framework that learns general, clinically informed features from brain MRIs. We trained this modular architecture on 49,246 individuals across 11 cohorts, using 17 diverse classification and regression tasks spanning cognition, clinical, diagnosis, demographics, and biomarkers. This yields aggregated, focused feature sets that capture rich, clinically- and biologically-relevant brain representations. We developed a sequential learning approach where tasks progressively build on previously learned representations. Through an analysis of 5,000 task sequences, we identified an optimal sequence length of six tasks and introduced a Donor Score metric to quantify each task's contribution to downstream performance. This analysis revealed five consistently strong donor tasks (Age, AD/MCI, MMSE, Hypertension, Hyperlipidemia) that formed the base of our sequential model. We demonstrated the utility of our learned representation, in various tasks beyond those included in the training set, to serve as the foundation for specialized secondary predictors. We further showed that using the learned feature representation can substantially increase the sample efficiency of secondary deep learning training tasks and models, as well as improve their accuracy.
V. Bashyam, G. Erus, Junhao Wen et al.· 0 citations
Alzheimer’s disease (AD) is a neurodegenerative disorder characterized by heterogeneous pathophysiological changes that begin years before symptoms emerge. Existing biomarkers like Aβ and pTau capture only fragments of this complexity, limiting diagnosis and therapeutic development. Leveraging high-resolution cerebrospinal fluid (CSF) proteomics, quantifying 2,492 proteins using tandem-mass-tag mass spectrometry (TMT-MS), in 1,104 ADNI participants, we identified pathways reflecting AD pathogenesis and stage-specific molecular events in-vivo. In biomarker-positive MCI (due-to-AD) and AD Dementia, beyond well-established metabolic and mitochondrial dysfunction, we observed upregulated neuropeptide signaling, G-protein-coupled receptors activity, and synaptic remodeling, highlighting underrecognized synaptic and signaling alterations. Asymptomatic AD showed significant alterations in mitochondrial metabolism, RNA processing, and extracellular matrix pathways. Across the continuum from asymptomatic AD to MCI (due-to-AD) and AD Dementia, 92 proteins were differentially abundant, revealing a stage-specific progression, with early disruptions in neurodevelopmental and extracellular vesicle-related pathways in asymptomatic and MCI (due-to-AD) participants, transitioning to impairments in intracellular signaling, synaptic architecture, and cytoskeletal integrity in AD Dementia. This progressive dysregulation supports a continuum model where early compensatory mechanisms gradually give way to widespread neuronal degeneration. Using machine learning, we derived CSF proteomic panels capable of accurately distinguishing disease stages (asymptomatic AD vs. MCI (due-to-AD): AUC = 0.92; MCI (due-to-AD) vs. AD Dementia: AUC = 0.87). In parallel, we developed machine learning models to estimate pathological burden (Aβ-PET, tau-PET), which substantially outperformed conventional biomarkers. These findings uncover protein signatures that reflect underlying AD biology and provide a foundation for stage-specific biomarkers and therapeutic targeting, with important implications for patient stratification and personalized intervention strategies.
Saima Rathore, E. Dammer, Anantharaman Shantaraman et al.· Molecular Neurodegeneration· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
