Accurate clinical toxicity prediction using multi-task deep neural nets and contrastive molecular explanations.

Explainable machine learning for molecular toxicity prediction is a promising approach for efficient drug development and chemical safety. A predictive ML model of toxicity can reduce experimental cost and time while mitigating ethical concerns by significantly reducing animal and clinical testing. Herein, we use a deep learning framework for simultaneously modeling in vitro, in vivo, and clinical toxicity data.

Weaker N-Terminal Interactions for the Protective over the Causative Aβ Peptide Dimer Mutants.

Knowing that abeta amyloid peptide (Aβ) dimers are the smallest and most abundant neurotoxic oligomers for Alzheimer's disease (AD), we used molecular simulations with advanced sampling methods (replica-exchange) to characterize and compare interactions between the N-termini (residues 1-16) of wild type (WT-WT) and five mutant dimers under constrained and unconstrained conditions. The number of contacts and distances between the N-termini, and contact maps of their conformational landscape illustrate substantial differences for a single residue change. The N-terminal contacts are significantly diminished for the dimers containing the monomers that protect against (WT-A2T) as compared with those that predispose toward (A2V-A2V) AD and for the control WT-WT dimers.

N-Terminal Hypothesis for Alzheimer's Disease.

Although the amyloid (abeta peptide, Aβ) hypothesis is 25 years old, is the dominant model of Alzheimer's disease (AD) pathogenesis, and guides the development of potential treatments, it is still controversial. One possible reason is a lack of a mechanistic path from the cleavage products of the amyloid precursor protein (APP) such as soluble Aβ monomer and soluble molecular fragments to the deleterious effects on synaptic form and function. From a review of the recent literature and our own published work including aggregation kinetics and structural morphology, Aβ clearance, molecular simulations, long-term potentiation measurements with inhibition binding, and the binding of a commercial monoclonal antibody, aducanumab, we hypothesize that the N-terminal domains of neurotoxic Aβ oligomers are implicated in causing the disease.

Unusual Morphologies of Poly(vinyl alcohol) Thin Films Adsorbed on Poly(dimethylsiloxane) Substrates.

Adsorption of poly(vinyl alcohol) (PVOH), 99% and 88% hydrolyzed poly(vinyl acetate), to poly(dimethylsiloxane) (PDMS) substrates was studied. The substrates were prepared by covalently attaching linear PDMS polymers of 2, 9, 17, 49, and 116 kDa onto silicon wafers. As the PDMS molecular weight/thickness increases, the adsorbed PVOH thin films progressively transition from continuous to discontinuous morphologies, including honeycomb and fractal/droplet.