A PET tracer is desired to help guide the discovery and development of disease-modifying therapeutics for neurodegenerative diseases characterized by neurofibrillary tangles (NFTs), the predominant tau pathology in Alzheimer disease (AD). We describe the preclinical characterization of the NFT PET tracer 18 F-MK-6240. Methods: In vitro binding studies were conducted with 3 H-MK-6240 in tissue slices and homogenates from cognitively normal and AD human brain donors to evaluate tracer affinity and selectivity for NFTs. Immunohistochemistry for phosphorylated tau was performed on human brain slices for comparison with 3 H-MK-6240 binding patterns on adjacent brain slices. PET studies were performed with 18 F-MK-6240 in monkeys to evaluate tracer kinetics and distribution in the brain. 18 F-MK-6240 monkey PET studies were conducted after dosing with unlabeled MK-6240 to evaluate tracer binding selectivity in vivo. Results: The 3 H-MK-6240 binding pattern was consistent with the distribution of phosphorylated tau in human AD brain slices. 3 H-MK-6240 bound with high affinity to human AD brain cortex homogenates containing abundant NFTs but bound poorly to amyloid plaque-rich, NFT-poor AD brain homogenates. 3 H-MK-6240 showed no displaceable binding in the subcortical regions of human AD brain slices and in the hippocampus/entorhinal cortex of non-AD human brain homogenates. In monkey PET studies, 18 F-MK-6240 displayed rapid and homogeneous distribution in the brain. The 18 F-MK-6240 volume of distribution stabilized rapidly, indicating favorable tracer kinetics. No displaceable binding was observed in self-block studies in rhesus monkeys, which do not natively express NFTs. Moderate defluorination was observed as skull uptake. Conclusion: 18 F-MK-6240 is a promising PET tracer for the in vivo quantification of NFTs in AD patients. Cur rently, the clinical evaluation of disease-modifying therapies for Alzheimer disease (AD) requires large, resourceintensive clinical trials focused on measuring cognitive endpoints, which are highly variable. A biomarker that could be used early in clinical development to build confidence in the ability of a therapeutic mechanism to modify disease progression would provide a valuable bridge to investment in a large efficacy study once adequate pharmacokinetics, safety, and tolerability have been established. Biomarkers currently in use (e.g., volumetric MRI, amyloid plaque PET, cerebrospinal fluid measures of amyloid-b and tau) either do not directly inform on modification of disease pathology (volumetric MRI) or do not correlate strongly enough with cognitive decline to measure therapeutic response (amyloid plaque PET and cerebrospinal fluid measures) (1,2). Therefore, there is an unmet need for sensitive biomarkers that quantify early pathologic changes and correlate closely to disease progression and clinical outcomes.Histologic analysis of brains from human autopsy cases have shown that the density and distribution of neurofibrillary tangles (NFTs) correlate with cognitive decline ...

The Absorption, Distribution, Metabolism, Elimination, and Toxicity (ADMET) properties of drug candidates are estimated to account for up to 50% of all clinical trial failures 1,2 . Predicting ADMET properties has therefore been of great interest to the cheminformatics and medicinal chemistry communities in recent decades. Traditional cheminformatics approaches, whether the learner is a random forest or a deep neural network, leverage fixed fingerprint feature representations of molecules. In contrast, in this paper, we learn the features most relevant to each chemical task at hand by representing each molecule explicitly as a graph, where each node is an atom and each edge is a bond. By applying graph convolutions to this explicit molecular representation, we achieve, to our knowledge, unprecedented accuracy in prediction of ADMET properties. By challenging our methodology with rigorous cross-validation procedures and prospective analyses, we show that deep featurization better enables molecular predictors to not only interpolate but also extrapolate to new regions of chemical space.

Neurofibrillary tangles (NFTs) made up of aggregated tau protein have been identified as the pathologic hallmark of several neurodegenerative diseases including Alzheimer's disease. In vivo detection of NFTs using PET imaging represents a unique opportunity to develop a pharmacodynamic tool to accelerate the discovery of new disease modifying therapeutics targeting tau pathology. Herein, we present the discovery of 6-(fluoro-(18)F)-3-(1H-pyrrolo[2,3-c]pyridin-1-yl)isoquinolin-5-amine, 6 ([(18)F]-MK-6240), as a novel PET tracer for detecting NFTs. 6 exhibits high specificity and selectivity for binding to NFTs, with suitable physicochemical properties and in vivo pharmacokinetics.

Given a particular descriptor/method combination, some quantitative structure–activity relationship (QSAR) datasets are very predictive by random-split cross-validation while others are not. Recent literature in modelability suggests that the limiting issue for predictivity is in the data, not the QSAR methodology, and the limits are due to activity cliffs. Here, we investigate, on in-house data, the relative usefulness of experimental error, distribution of the activities, and activity cliff metrics in determining how predictive a dataset is likely to be. We include unmodified in-house datasets, datasets that should be perfectly predictive based only on the chemical structure, datasets where the distribution of activities is manipulated, and datasets that include a known amount of added noise. We find that activity cliff metrics determine predictivity better than the other metrics we investigated, whatever the type of dataset, consistent with the modelability literature. However, such metrics cannot distinguish real activity cliffs due to large uncertainties in the activities. We also show that a number of modern QSAR methods, and some alternative descriptors, are equally bad at predicting the activities of compounds on activity cliffs, consistent with the assumptions behind “modelability.” Finally, we relate time-split predictivity with random-split predictivity and show that different coverages of chemical space are at least as important as uncertainty in activity and/or activity cliffs in limiting predictivity.

Zileuton, an inhibitor of 5-lipooxygenase, the initial enzyme in the leukotriene pathway, was marketed as a new treatment for asthma. This drug has been associated with liver toxicity, which has limited its clinical usefulness. We provide evidence here that the liver toxicity likely involves a sequence of biotransformations leading to 2-acetylbenzothiophene (2-ABT), which is subsequently metabolized to give a reactive intermediate(s). In vitro experiments with the human lymphoblast MCL5 cell line demonstrated that 2-ABT is cytotoxic in a P450-dependent manner. Human liver microsome (HLM) incubations with 2-ABT revealed the formation of two short-lived oxidized species, "M + 16" and "M + 32". Both of these metabolites formed adducts in the presence of GSH or NAC. Singly oxidized M + 16 adducts, from either GSH or NAC, appeared to be unstable in acidic medium and eliminated water readily to form a new compound. Authentic synthetic standards demonstrated that 2-ABT-S-oxide M1 corresponded to the M + 16 metabolite and that the S-oxide underwent nucleophilic addition with GSH and NAC to produce the singly oxidized adducts observed in HLM. The S-oxide adducts readily eliminated water to form a rearomatized 2-ABT-GSH adduct or 2-ABT-NAC adduct. Coelution experiments with the synthetic standard confirmed the structure of the eliminated 2-ABT-NAC adduct C1. LC/MS analyses of urine samples collected from rats dosed with zileuton indicate that C1 is a metabolite of zileuton formed in vivo. The in vitro and in vivo data presented here demonstrate the formation of 2-ABT from zileuton and its further bioactivation to a potentially toxic metabolite.