Hjelmlundberg3518

Advances in deep neural network (DNN) based molecular property prediction have recently led to the development of models of remarkable accuracy and generalization ability, with graph convolutional neural networks (GCNNs) reporting state-of-the-art performance for this task. However, some challenges remain and one of the most important that needs to be fully addressed concerns uncertainty quantification. DNN performance is affected by the volume and the quality of the training samples. Therefore, establishing when and to what extent a prediction can be considered reliable is just as important as outputting accurate predictions, especially when out-of-domain molecules are targeted. Recently, several methods to account for uncertainty in DNNs have been proposed, most of which are based on approximate Bayesian inference. Among these, only a few scale to the large datasets required in applications. Evaluating and comparing these methods has recently attracted great interest, but results are generally fragmented and absent for molecular property prediction. In this paper, we aim to quantitatively compare scalable techniques for uncertainty estimation in GCNNs. We introduce a set of quantitative criteria to capture different uncertainty aspects, and then use these criteria to compare MC-Dropout, Deep Ensembles, and bootstrapping, both theoretically in a unified framework that separates aleatoric/epistemic uncertainty and experimentally on public datasets. Our experiments quantify the performance of the different uncertainty estimation methods and their impact on uncertainty-related error reduction. Our findings indicate that Deep Ensembles and bootstrapping consistently outperform MC-Dropout, with different context-specific pros and cons. Our analysis leads to a better understanding of the role of aleatoric/epistemic uncertainty, also in relation to the target dataset features, and highlights the challenge posed by out-of-domain uncertainty.One of the key requirements for incorporating machine learning (ML) into the drug discovery process is complete traceability and reproducibility of the model building and evaluation process. With this in mind, we have developed an end-to-end modular and extensible software pipeline for building and sharing ML models that predict key pharma-relevant parameters. The ATOM Modeling PipeLine, or AMPL, extends the functionality of the open source library DeepChem and supports an array of ML and molecular featurization tools. We have benchmarked AMPL on a large collection of pharmaceutical data sets covering a wide range of parameters. Our key findings indicate that traditional molecular fingerprints underperform other feature representation methods. We also find that data set size correlates directly with prediction performance, which points to the need to expand public data sets. Uncertainty quantification can help predict model error, but correlation with error varies considerably between data sets and model types. Our findings point to the need for an extensible pipeline that can be shared to make model building more widely accessible and reproducible. This software is open source and available at https//github.com/ATOMconsortium/AMPL.Acquired drug resistance in epidermal growth factor receptor (EGFR) mutant non-small-cell lung cancer is a persistent challenge in cancer therapy. Previous studies of trisubstituted imidazole inhibitors led to the serendipitous discovery of inhibitors that target the drug resistant EGFR(L858R/T790M/C797S) mutant with nanomolar potencies in a reversible binding mechanism. To dissect the molecular basis for their activity, we determined the binding modes of several trisubstituted imidazole inhibitors in complex with the EGFR kinase domain with X-ray crystallography. These structures reveal that the imidazole core acts as an H-bond acceptor for the catalytic lysine (K745) in the "αC-helix out" inactive state. Selective N-methylation of the H-bond accepting nitrogen ablates inhibitor potency, confirming the role of the K745 H-bond in potent, noncovalent inhibition of the C797S variant. Insights from these studies offer new strategies for developing next generation inhibitors targeting EGFR in non-small-cell lung cancer.Microketides A and B (1 and 2), a pair of new C-11 epimeric polyketides, were obtained from the gorgonian-derived fungus Microsphaeropsis sp. RA10-14 collected from the South China Sea. The absolute configurations of 1 and 2 were assigned by the modified Mosher's method, TDDFT-ECD, and NMR calculations. Compounds 1 and 2 were evaluated for antibacterial, antifungal, and growth inhibition of marine phytoplankton activities. Microketide A (1) exhibited promising inhibitory activity against Pseudomonas aeruginosa, Nocardia brasiliensis, Kocuria rhizophila, and Bacillus anthraci with the same MIC value as ciprofloxacin (0.19 μg/mL).Molecular mechanics force fields have been shown to differ in their predictions of biomolecular processes such as protein folding. To test how force field differences affect predicted polypeptide behavior, we created a mechanically perturbed model of the β-stranded I91 titin domain with the A-strand detached from the fold based on atomic force spectroscopy data and examined its refolding behavior using six different force fields. We found that different force fields varied significantly in their ability to refold the mechanically perturbed I91 domain. Examination of the perturbed I91 unfolded state revealed that all five Amber force fields over-sample a specific region of the Ramachandran plot thereby creating unfolded state intermediates which are not predicted by the Charmm 22* force field. Simulations of perturbed I91 refolding with Amber FB15 revealed that Amber FB15 destabilizes stable portions of I91 thereby contradicting experimental stability analyses. Finally, inspection of the perturbed I91 unfolded state along with equilibration simulations of the Ac-(AAQAA)3-NH2 peptide suggest that high dihedral torsional barriers cause the Amber ff14SB force field to predict higher helical lifetimes relative to other force fields. These results suggest that using mechanically perturbed models can provide a controlled method to gain insight into how force fields affect polypeptide behavior.Quantum-mechanical/molecular-mechanical (QM/MM) methods are essential to the study of metalloproteins, but the relative importance of sampling and degree of QM treatment in achieving quantitative predictions is poorly understood. We study the relative magnitude of configurational and QM-region sensitivity of energetic and electronic properties in a representative Zn2+ metal binding site of a DNA methyltransferase. To quantify property variations, we analyze snapshots extracted from 250 ns of molecular dynamics simulation. To understand the degree of QM-region sensitivity, we perform analysis using QM regions ranging from a minimal 49-atom region consisting only of the Zn2+ metal and its four coordinating Cys residues up to a 628-atom QM region that includes residues within 12 Å of the metal center. Over the configurations sampled, we observe that illustrative properties (e.g., rigid Zn2+ removal energy) exhibit large fluctuations that are well captured with even minimal QM regions. Nevertheless, for both energetic and electronic properties, we observe a slow approach to asymptotic limits with similarly large changes in absolute values that converge only with larger (ca. 300-atom) QM region sizes. For the smaller QM regions, the electronic description of Zn2+ binding is incomplete the metal binds too tightly and is too stabilized by the strong electrostatic potential of MM point charges, and the Zn-S bond covalency is overestimated. Overall, this work suggests that efficient sampling with QM/MM in small QM regions is an effective method to explore the influence of enzyme structure on target properties. At the same time, accurate descriptions of electronic and energetic properties require a larger QM region than the minimal metal-coordinating residues in order to converge treatment of both metal-local bonding and the overall electrostatic environment.The environmental impacts of packaging and food service ware (FSW) are increasingly the subject of government policy, public discourse, and industry commitments. While some consideration is given to reducing the impacts of packaging across its entire life cycle, most of the focus is on packaging waste or feedstock substitution. Efforts typically focus on specific packaging characteristics, or material attributes, commonly perceived to be environmentally preferable. This article summarizes an extensive meta-review of existing published literature that was performed to determine whether the material attributes recyclability, recycled content, compostability, and biobased, commonly considered to be environmentally beneficial, correlate with lower net environmental impacts across the full life cycle of the packaging and FSW. Seventy-one unique life cycle assessment (LCA) studies that quantify the environmental impacts throughout the entire life cycle of packaging and FSW were analyzed. These studies included over 5000 comparisons for 13 impact categories commonly analyzed in LCA studies. The results from the meta-review identified a number of instances where material attributes do not correlate with environmental benefits for packaging and FSW. Rather, other characteristics such as material choice or mass of the packaging/FSW products can have higher influence in determining life cycle impacts.Ionotropic γ-aminobutyric acid (GABA) receptors (GABARs) represent an important insecticide target. Currently used GABAR-targeting insecticides are non-competitive antagonists (NCAs) of these receptors. Recent studies have demonstrated that competitive antagonists (CAs) of GABARs have functions of inhibiting insect GABARs similar to NCAs and that they also exhibit insecticidal activity. CAs have different binding sites and different mechanisms of action compared to those of NCAs. Therefore, GABAR CAs should have the potential to be developed into novel insecticides, which could be used to overcome the developed resistance of insect pests to conventional NCA insecticides. Although research on insect GABAR CAs has lagged behind that on mammalian GABAR CAs, research on the CAs of insect ionotropic GABARs has made great progress in recent years, and several series of heterocyclic compounds, such as 3-isoxazolols and 6-iminopyridazines, have been identified as insect GABAR CAs. In this review, we briefly summarize the design strategies, structures, and biological activities of the novel GABAR CAs that have been found in the past decade. Updated information about GABAR CAs may benefit the design and development of novel GABAR-targeting insecticides.The usual understanding in polymer electrolyte design is that an increase in the polymer dielectric constant results in reduced ion aggregation and therefore increased ionic conductivity. We demonstrate here that in a class of polymers with extensive metal-ligand coordination and tunable dielectric properties, the extent of ionic aggregation is delinked from the ionic conductivity. The polymer systems considered here comprise ether, butadiene, and siloxane backbones with grafted imidazole side-chains, with dissolved Li+, Cu2+, or Zn2+ salts. The nature of ion aggregation is probed using a combination of X-ray scattering, electron paramagnetic resonance (in the case where the metal cation is Cu2+), and polymer field theory-based simulations. Polymers with less polar backbones (butadiene and siloxane) show stronger ion aggregation in X-ray scattering compared to those with the more polar ether backbone. The Tg-normalized ionic conductivities were however unaffected by the extent of aggregation. The results are explained on the basis of simulations which indicate that polymer backbone polarity does impact the microstructure and the extent of ion aggregation but does not impact percolation, leading to similar ionic conductivity regardless of the extent of ion aggregation.