Equivariant Graph Neural Networks for Toxicity Prediction.

Predictive modeling of toxicity is a crucial step in the drug discovery pipeline. It can help filter out molecules with a high probability of failing in the early stages of de novo drug design. Thus, several machine learning (ML) models have been developed to predict the toxicity of molecules by combining classical ML techniques or deep neural networks with well-known molecular representations such as fingerprints or 2D graphs. But the more natural, accurate representation of molecules is expected to be defined in physical 3D space like in ab initio methods. Recent studies successfully used equivariant graph neural networks (EGNNs) for representation learning based on 3D structures to predict quantum-mechanical properties of molecules. Inspired by this, we investigated the performance of EGNNs to construct reliable ML models for toxicity prediction. We used the equivariant transformer (ET) model in TorchMD-NET for this. Eleven toxicity data sets taken from MoleculeNet, TDCommons, and ToxBenchmark have been considered to evaluate the capability of ET for toxicity prediction. Our results show that ET adequately learns 3D representations of molecules that can successfully correlate with toxicity activity, achieving good accuracies on most data sets comparable to state-of-the-art models. We also test a physicochemical property, namely, the total energy of a molecule, to inform the toxicity prediction with a physical prior. However, our work suggests that these two properties can not be related. We also provide an attention weight analysis for helping to understand the toxicity prediction in 3D space and thus increase the explainability of the ML model. In summary, our findings offer promising insights considering 3D geometry information via EGNNs and provide a straightforward way to integrate molecular conformers into ML-based pipelines for predicting and investigating toxicity prediction in physical space. We expect that in the future, especially for larger, more diverse data sets, EGNNs will be an essential tool in this domain.

Scalable Synthesis of Carbon Nanomembranes from Amorphous Molecular Layers.

Nanoporous carbon nanomembranes (CNMs) created by self-assembled monolayers ideally combine a high water flux and precise ion selectivity for molecular separation and water desalination. However, their practical implementation is often challenged by the availability of large epitaxial substrates, limiting the membrane up-scaling. Here, we report a scalable synthesis of CNMs from poly(4-vinylbiphenyl) (PVBP) spin-coated on SiO2/Si wafers. Electron irradiation of the amorphous PVBP molecular layers induces the formation of a continuous membrane with a thickness of 15 nm and a high density of subnanometer pores, providing a water permeance as high as 530 L m-2 h-1 bar-1, while repelling ions and molecules larger than 1 nm in size. A further introduction of a reinforced porous block copolymer layer enables the fabrication of centimeter-scale CNM composites that efficiently separate organic dyes from water. These results suggest a feasible route for large-scale nanomembrane fabrication.

Aging of Industrial Polypropylene Surfaces in Detergent Solution and Its Consequences for Biofilm Formation.

The performance of plastic components in water-bearing parts of industrial and household appliances, often in the presence of harsh environments and elevated temperatures, critically relies on the mechanical and thermal polymer stability. In this light, the precise knowledge of aging properties of polymers formulated with dedicated antiaging additive packages as well as various fillers is crucial for long-time device warranty. We investigated and analysed the time-dependent, polymer-liquid interface aging of different industrial performance polypropylene samples in aqueous detergent solution at high temperatures (95 °C). Special emphasis was put on the disadvantageous process of consecutive biofilm formation that often follows surface transformation and degradation. Atomic force microscopy, scanning electron microscopy, and infrared spectroscopy were used to monitor and analyse the surface aging process. Additionally, bacterial adhesion and biofilm formation was characterised by colony forming unit assays. One of the key findings is the observation of crystalline, fibre-like growth of ethylene bis stearamide (EBS) on the surface during the aging process. EBS is a widely used process aid and lubricant enabling the proper demoulding of injection moulding plastic parts. The aging-induced surface-covering EBS layers changed the surface morphology and promoted bacterial adhesion as well as biofilm formation of Pseudomonas aeruginosa.

Pd Nanoparticle-Loaded Smart Microgel-Based Membranes as Reusable Catalysts.

In this work, palladium-loaded smart membranes made by UV cross-linking of thermoresponsive microgels are prepared to obtain a reusable, catalytically active material which can, for example, be implemented in chemical reactors. The membranes are examined with respect to their coverage of a supporting mesh via atomic force microscopy measurements. Force indentation mapping was performed in the dried, collapsed state and in the swollen state in water to determine the Young modulus. Furthermore, we compare the catalytic activity of the membrane with the corresponding suspended colloidal nanoparticle microgel hybrids. For this purpose, the reduction of 4-nitrophenol is an established model reaction to quantify the catalytic activity by UV-vis spectroscopy. The membrane is embedded inside a continuous stirred tank reactor equipped for continuous monitoring of the reaction progress. Although catalysis with membranes shows lower catalytic activity than freely dispersed particles, membranes allow straightforward separation and recycling of the catalyst. The fabricated membranes in this work show no decrease in catalytic activity between several cycles, unlike free particles. The feasible and durable deposition of catalytically active inter-cross-linked microgel particles on commercial nylon meshes as supporting scaffolds, as demonstrated in this work, is promising for up-scaling of continuous industrial processes.

Magnetron Sputtering of Transition Metals as an Alternative Production Means for Antibacterial Surfaces.

In the light of the SARS-CoV-2 pandemic and growing numbers of bacteria with resistance to antibiotics, the development of antimicrobial coatings is rising worldwide. Inorganic coatings are attractive because of low environmental leakage and wear resistance. Examples for coatings are hot metal dipping or physical vapor deposition of nanometer coatings. Here, magnetron sputtering of various transition metals, such as gold, ruthenium and tantalum, was investigated. Metal films were characterized by scanning electron microscopy (SEM), atomic force microscopy (AFM) and energy dispersive X-ray spectroscopy (EDX). We investigated the growth of Pseudomonas aeruginosa isolated from household appliances on different sputter-coated metal surfaces. The fine-grained nanometric structure of these metal coatings was between 14 nm (tantalum) and 26 nm (gold) and the roughness was in a range of 164 pm (ruthenium) to 246 pm (gold). Antibacterial efficacy of metal surfaces followed the order: gold > tantalum > ruthenium. Interestingly, gold had the strongest inhibitory effect on bacterial growth, as analyzed by LIVE/DEAD and CFU assay. High-magnification SEM images showed dead bacteria characterized by shrinkage induced by metal coatings. We conclude that sputtering might be a new application for the development of antimicrobial surfaces on household appliances and or surgical instruments.

Thickness-Varied Carbon Nanomembranes from Polycyclic Aromatic Hydrocarbons.

Despite the prospects of intrinsically porous planar nanomaterials in separation applications, their synthesis on a large scale remains challenging. In particular, preparing water-selective carbon nanomembranes (CNMs) from self-assembled monolayers (SAMs) is limited by the cost of epitaxial metal substrates and molecular precursors with specific chemical functionalities. In this work, we present a facile fabrication of CNMs from polycyclic aromatic hydrocarbons (PAHs) that are drop-cast onto arbitrary supports, including foils and metalized films. The electron-induced carbonization is shown to result in continuous membranes of variable thickness, and the material is characterized with a number of spectroscopic and microscopic techniques. Permeation measurements with freestanding membranes reveal a high degree of porosity, but the selectivity is found to strongly depend on the thickness. While the permeance of helium remains almost the same for 6.5 and 3.0 nm thick CNMs, water permeance increases by 2 orders of magnitude. We rationalize the membrane performance with the help of kinetic modeling and vapor adsorption experiments.

Thermoresponsive Microgel-Based Free-Standing Membranes: Influence of Different Microgel Cross-Linkers on Membrane Function.

In this study we show a possibility to produce thermoresponsive, free-standing microgel membranes based on N-isopropylacrylamide (NIPAM) and the UV-sensitive comonomer 2-hydroxy-4-(methacryloyloxy)benzophenone (HMABP). To influence the final network structure and functionality of the membranes, we use different cross-linkers in the microgel syntheses and characterize the resulting structural microgel properties and the swelling behavior by means of AFM, FTIR, and PCS measurements. Varying the cross-linker results in significant changes in the structure and swelling behavior of the individual microgels and has an influence on the incorporation of the comonomer, which is essential for subsequent photochemical membrane formation. We investigate the ion transport through the different membranes by temperature-dependent resistance measurements revealing a sharp increase in resistance when the copolymer microgels reach their collapsed state. The resistance of the membranes can be adjusted by different cross-linkers and the associated incorporation of the comonomer. Furthermore, we show that transferring a reversible cross-linker from a cross-linked state to an un-cross-linked state strongly influences the membrane properties and even reverses the switching behavior, while the mechanical stability of the membrane is maintained.