Objective: Understanding the heterogeneous pathology in Alzheimer's disease and related tauopathies is one of the most urgent and fundamental challenges facing the discovery of novel disease-modifying therapies. Through monitoring ensembles of toxic and nontoxic tau oligomers spontaneously formed in cells, our biosensor technology can identify tool compounds that modulate tau oligomer structure and toxicity, providing much needed insight into the nature and properties of toxic tau oligomers. Background: Tauopathies are a group of neurodegenerative disorders characterized by pathologic aggregation of the microtubule binding protein tau. Recent studies suggest that tau oligomers are the primary toxic species in tauopathies. New/Updated Hypothesis: We hypothesize that tau biosensors capable of monitoring tau oligomer conformation are able to identify tool compounds that modulate the structure and conformation of these tau assemblies, providing key insight into the unique structural fingerprints of toxic tau oligomers. These fingerprints will provide gravely needed biomarker profiles to improve staging of early tauopathy pathology and generate lead compounds for potential new therapeutics. Our time-resolved fluorescence resonance energy transfer biosensors provide us an exquisitely sensitive technique to monitor minute structural changes in monomer and oligomer conformation. In this proof-ofconcept study, we identified a novel tool compound, MK-886, which directly binds tau, perturbs the conformation of toxic tau oligomers, and rescues tau-induced cytotoxicity. Furthermore, we show that MK-886 alters the conformation of tau monomer at the proline-rich and microtubule binding regions, stabilizing an on-pathway oligomer. Major Challenges for the Hypothesis: Our approach monitors changes in the ensemble of assemblies that are spontaneously formed in cells but does not specifically isolate or enrich unique toxic tau species. However, time-resolved fluorescence resonance energy transfer does not provide highresolution, atomic scale information, requiring additional experimental techniques to resolve the structural features stabilized by different tool compounds. David D. Thomas holds equity in and serves as an executive officer for Photonic Pharma LLC, a company that owns intellectual property related to technology used in part of this project. These relationships have been reviewed and managed by the University of Minnesota in accordance with its conflict-of-interest polices.

No abstract

In this work, we improve the performance of multi-atlas segmentation (MAS) by integrating the recently proposed VoteNet model with the joint label fusion (JLF) approach. Specifically, we first illustrate that using a deep convolutional neural network to predict atlas probabilities can better distinguish correct atlas labels from incorrect ones than relying on image intensity difference as is typical in JLF. Motivated by this finding, we propose VoteNet+, an improved deep network to locally predict the probability of an atlas label to differ from the label of the target image. Furthermore, we show that JLF is more suitable for the VoteNet framework as a label fusion method than plurality voting. Lastly, we use Platt scaling to calibrate the probabilities of our new model. Results on LPBA40 3D MR brain images show that our proposed method can achieve better performance than VoteNet.

No abstract