Molecular Systems Design & Engineering最新文献

Graph neural networks (GNN) have been demonstrated to correlate molecular structure with properties, enabling rapid evaluation of molecules for a given application. Molecular properties, including ground and excited states, are crucial to analyzing molecular behavior. However, while attention-based mechanisms and pooling methods have been optimized to accurately predict specific properties, no versatile models can predict diverse molecular properties. Here, we present graph neural networks that predict a wide range of properties with high accuracy. Model performance is high regardless of dataset size and origin. Further, we demonstrate an implementation of hierarchical pooling enabling high-accuracy prediction of excited state properties by effectively weighing aspects of features that correlate better with target properties. We show that graph attention networks consistently outperform convolution networks and linear regression, particularly for small dataset sizes. The graph attention model is more accurate than previous message-passing neural networks developed for the prediction of diverse molecular properties. Hence, the model is an efficient tool for screening and designing molecules for applications that require tuning multiple molecular properties.

Block copolymer (BCP) self-assembly leads to nanostructured materials with diverse ordered morphologies, some of which are attractive for transport applications. Multiblock AB copolymers are of interest as they offer a larger design parameter space than diblock copolymers allowing researchers to tailor their self-assembly to achieve target morphologies. In this study, we investigate the phase behavior of symmetric A_xB_yA_zB_yA_x and B_xA_yB_zA_yB_x pentablock copolymers (pentaBCPs) where A and B monomers have the same statistical segment length. We use a combination of self-consistent field theory (SCFT) calculations and molecular dynamics (MD) simulations to link the polymer design parameters, namely the fraction of middle block volume to the volume of all blocks of same type, τ, overall volume fraction of A block, f_A, and segregation strength, χN, to the equilibrium morphologies and the distributions of chain conformations in these morphologies. In the phase diagrams calculated using SCFT, we observe broader double gyroid windows and the existence of lamellar morphologies even at small values f_A in contrast to what has been seen for diblock copolymers. We also see a reentrant phase sequence of double gyroid → cylinder → lamellae → cylinder → double gyroid with increasing τ at fixed f_A. The chain conformations adopted in these morphologies are sampled in coarse-grained MD simulations and quantified with distributions of the chain end-to-end distance and fractions of chains whose middle (A or B) and end (A or B) blocks remain within domains of same chemistry (A or B). These analyses show that the pentaBCP chains adopt “looping”, “bridging”, and “hybrid” (both looping and bridging) conformations, with a majority of the chains adopting the hybrid conformation. The spatial distributions for each of the blocks in the pentaBCPs show that blocks of the same type in a chain locally segregate within the same domains, with shorter blocks segregating towards the domain boundaries and longer blocks filling the domain interior. This combined SCFT-MD approach enables us to rapidly screen the extensive pentaBCP design space to identify design rules for transport-favorable morphologies as well as verify the chain conformations and spatial arrangements associated with the theory predicted reentrant phase behavior.

Structural proteins like silk, squid ring teeth, elastin, collagen, or resilin, among others, are inspiring the development of new sustainable biopolymeric materials for applications including healthcare, food, soft robotics, or textiles. Furthermore, advances in the fields of soft materials and synthetic biology have a joint great potential to guide the design of novel structural proteins, despite both fields progressing mostly in a separate fashion so far. Using recombinant DNA technologies and microbial fermentations, we can design new structural proteins with monomer-level sequence control and a dispersity of ca. 1.0, based on permutations of tandem repeats derived from natural structural proteins. However, the molecular design of recombinant and repetitive structural proteins is a nontrivial task that is generally approached using low-throughput trial-and-error experimentation. Here, we review recent progress in this area, in terms of structure–function relationships and DNA synthesis technologies. We also discuss experimental and computational advances towards the establishment of rapid prototyping pipelines for this family of biopolymers. Finally, we highlight future challenges to make protein-based materials a commercially viable alternative to current fossil-based polymers.

Peptides are naturally potent and selective therapeutics with massive potential; however, low cell membrane permeability limits their clinical implementation, particularly for hydrophilic, anionic peptides with intracellular targets. To overcome this limitation, esterification of anionic carboxylic acids on therapeutic peptides can simultaneously increase hydrophobicity and net charge to facilitate cell internalization, whereafter installed esters can be cleaved hydrolytically to restore activity. To date, however, most esterified therapeutics contain either a single esterification site or multiple esters randomly incorporated on multiple sites. This investigation provides molecular engineering insight into how the number and position of esters installed onto the therapeutic peptide α carboxyl terminus 11 (αCT11, RPRPDDLEI) with 4 esterification sites affect hydrophobicity and the hydrolysis process that reverts the peptide to its original form. After installing methyl esters onto αCT11 using Fischer esterification, we isolated 5 distinct products and used 2D nuclear magnetic resonance spectroscopy, reverse-phase high performance liquid chromatography, and mass spectrometry to determine which residues were esterified in each and the resulting increase in hydrophobicity. We found esterifying the C-terminal isoleucine to impart the largest increase in hydrophobicity. Monitoring ester hydrolysis showed the C-terminal isoleucine ester to be the most hydrolytically stable, followed by the glutamic acid, whereas esters on aspartic acids hydrolyze rapidly. LC-MS revealed the formation of transient intramolecular aspartimides prior to hydrolysis to carboxylic acids. In vitro proof-of-concept experiments showed esterifying αCT11 to increase cell migration into a scratch, highlighting the potential of multi-site esterification as a tunable, reversible strategy to enable the delivery of therapeutic peptides.

Metal–organic frameworks (MOFs) are promising platforms for designing photoresponsive materials due to their structural versatility and tunable properties. However, challenges remain in fine-tuning the photoresponsive behavior while maintaining the high stability of MOFs. In this study, we synthesized a MOF containing redox-active pyromellitic diimide (PMDI) groups and unsaturated Zr₆ clusters named Zr-PMDI-DMF and fine-tuned its photochromic properties by exchanging the coordination solvent molecules on the Zr sites. Unlike traditional Zr₆ clusters with bidentate carboxylate coordination, Zr-PMDI-DMF features monodentate carboxylate coordination with the exposed Zr sites occupied by solvent molecules. We post-synthetically exchanged the coordinated N,N-dimethylformamide (DMF) solvent molecules with 2-(dimethylamino)ethanol (DMAE), N-methyltetrahydropyrrole (NMP), and dimethyl sulfoxide (DMSO) and determined the structures of the coordinated solvent molecules using single-crystal X-ray diffraction. Through photochromic and bleaching cycle experiments, electron paramagnetic resonance spectroscopy, and density functional theory calculations, we found that the coordinated solvents act as electron donors. In contrast, PMDI ligands act as electron acceptors, causing intra-framework electron transfer and photochromism. The rate of the photochromic response correlated with the electron-donating ability of the solvents, following the trend of DMAE > NMP > DMSO > DMF.

Developing an accurate predictive model of solvent effects on reaction kinetics is a challenging task, yet it can play an important role in process development. While first-principles or machine learning models are often compute- or data-intensive, simple surrogate models, such as multivariate linear or quadratic regression models, are useful when computational resources and data are scarce. The judicious choice of a small set of training data, i.e., a set of solvents in which quantum mechanical (QM) calculations of liquid-phase rate constants are to be performed, is critical to obtaining a reliable model. This is, however, made especially challenging by the highly irregular shape of the discrete space of possible experiments (solvent choices). In this work, we demonstrate that when choosing a set of computer experiments to generate training data, the D-optimality criterion value of the chosen set correlates well with the likelihood of achieving good model performance. With the Menshutkin reaction of pyridine and phenacyl bromide as a case study, this finding is further verified via the evaluation of the surrogate models regressed using D-optimal solvent sets generated from four distinct selection spaces. We also find that incorporating quadratic terms in the surrogate model and choosing the D-optimal solvent set from a selection space similar to the test set can significantly improve the accuracy of reaction rate constant predictions while using a small training dataset. Our approach holds promise for the use of statistical optimality criteria for other types of computer experiments, supporting the construction of surrogate models with reduced resource and data requirements.

Biomineralization has garnered attention not only for its fundamental role in understanding the mechanisms of biomineral formation but also as a method for fabricating next-generation functional materials. In this study, we investigated the nucleation, crystal growth, and particle growth processes of calcium phosphates (CaPs) formed using selective mineralization at the hydrogel interface induced by the fusion of peptide hydrogels. After 1 day of mineralization, band-like white precipitates were observed at the fusion interface of the hydrogels. Notably, the nucleation and crystal growth of the mineralized CaP exhibited different behaviors owing to the differences in the properties of the reaction interface for mineralization. The selective nucleation and crystal growth of the CaPs at the hydrogel interface were attributed to (1) the local concentration of mineral sources near the peptide network, driven by electrostatic interactions between the polar functional groups and mineral source ions, and (2) selective crystal growth of the CaPs induced by the nanostructure of the surface functional groups.

Multifunctional composites with rapid self-healing performance have been widely applied in various fields. However, different types of fillers result in decreased self-healing efficiency and present agglomeration and poor compatibility especially at high filler contents. Here, based on the different surface modifications of barium titanate (BT) and silicon carbide (SiC) and the amide-bond synergistic effects between these fillers, self-healing supramolecular composites with high filler contents (up to 30%) are reported, and exhibit high strength, dielectric and thermal-conduction properties. Modification significantly improves the dispersion of these fillers, and greatly enhances the coexistence and synergy between these fillers. This three-phase amide-bonded supramolecular composite exhibits a high tensile strength of 3.22 MPa compared to other self-healing materials such as self-healing hydrogels, a high dielectric constant of 23, a high thermal conductivity of 0.36 W m⁻¹ K⁻¹ and a superior self-healing efficiency of above 94%. These performances are ascribed to the formation of amide bonds between the amino groups in 3-aminopropyltriethoxysilane (KH550)-modified silicon carbide (SiC-NH₂) and the carboxyl groups in tartaric acid (TA)-modified barium titanate (BT-TA), which can provide efficient supramolecular interactions between different fillers, as well as more reversible hydrogen bonding for the matrix. This three-phase amide-bonded supramolecular composite provides an effective strategy to improve the self-healing properties of multifunctional composites, and will bring pioneering functions to electronic packaging materials, dielectric energy storage materials, environmental energy and other fields, which can open up broad application prospects.

The estimation of polymer properties is of crucial importance in many domains such as energy, healthcare, and packaging. Recently, graph neural networks (GNNs) have shown promising results for the prediction of polymer properties based on supervised learning. However, the training of GNNs in a supervised learning task demands a huge amount of polymer property data that is time-consuming and computationally/experimentally expensive to obtain. Self-supervised learning offers great potential to reduce this data demand through pre-training the GNNs on polymer structure data only. These pre-trained GNNs can then be fine-tuned on the supervised property prediction task using a much smaller labeled dataset. We propose to leverage self-supervised learning techniques in GNNs for the prediction of polymer properties. We employ a recent polymer graph representation that includes essential features of polymers, such as monomer combinations, stochastic chain architecture, and monomer stoichiometry, and process the polymer graphs through a tailored GNN architecture. We investigate three self-supervised learning setups: (i) node- and edge-level pre-training, (ii) graph-level pre-training, and (iii) ensembled node-, edge- & graph-level pre-training. We additionally explore three different transfer strategies of fully connected layers with the GNN architecture. Our results indicate that the ensemble node-, edge- & graph-level self-supervised learning with all layers transferred depicts the best performance across dataset size. In scarce data scenarios, it decreases the root mean square errors by 28.39% and 19.09% for the prediction of electron affinity and ionization potential compared to supervised learning without the pre-training task.

Electrowetting display (EWD) technology is among the most promising reflective display technologies due to its full-color capabilities and fast video-speed performance. The colored EWD inks are typically prepared by dissolving soluble organic dyes in non-polar solvents, which significantly influence the color performance, electro-optical behaviour, and longevity of EWD devices. In this study, density functional theory (DFT) at the PBE1PBE/6-31G* level and time-dependent density functional theory (TD-DFT) at the M06-2X/6-31G* level were utilized to calculate a series of benzobisthiadiazole-based donor–acceptor–donor (D–A–D) type near-infrared organic dyes for EWDs, providing structural and spectral data to aid in spectral assignment. The quantum chemical calculations' results align with our experimental synthesis data, showing molecular colors spanning blue, green, and cyan. Detailed investigations into the properties of these dyes, including absorption, electro-optical response, and photo-stability, were conducted. The experimental outcomes indicate that these organic dyes are excellent candidates for EWD applications.