When Do Quantum Mechanical Descriptors Help Graph Neural Networks to Predict Chemical Properties?

Deep graph neural networks are extensively utilized to predict chemical reactivity and molecular properties. However, because of the complexity of chemical space, such models often have difficulty extrapolating beyond the chemistry contained in the training set. Augmenting the model with quantum mechanical (QM) descriptors is anticipated to improve its generalizability. However, obtaining QM descriptors often requires CPU-intensive computational chemistry calculations. To identify when QM descriptors help graph neural networks predict chemical properties, we conduct a systematic investigation of the impact of atom, bond, and molecular QM descriptors on the performance of directed message passing neural networks (D-MPNNs) for predicting 16 molecular properties. The analysis surveys computational and experimental targets, as well as classification and regression tasks, and varied data set sizes from several hundred to hundreds of thousands of data points. Our results indicate that QM descriptors are mostly beneficial for D-MPNN performance on small data sets, provided that the descriptors correlate well with the targets and can be readily computed with high accuracy. Otherwise, using QM descriptors can add cost without benefit or even introduce unwanted noise that can degrade model performance. Strategic integration of QM descriptors with D-MPNN unlocks potential for physics-informed, data-efficient modeling with some interpretability that can streamline de novo drug and material designs. To facilitate the use of QM descriptors in machine learning workflows for chemistry, we provide a set of guidelines regarding when and how to best leverage QM descriptors, a high-throughput workflow to compute them, and an enhancement to Chemprop, a widely adopted open-source D-MPNN implementation for chemical property prediction.

Chemprop: A Machine Learning Package for Chemical Property Prediction.

Deep learning has become a powerful and frequently employed tool for the prediction of molecular properties, thus creating a need for open-source and versatile software solutions that can be operated by nonexperts. Among the current approaches, directed message-passing neural networks (D-MPNNs) have proven to perform well on a variety of property prediction tasks. The software package Chemprop implements the D-MPNN architecture and offers simple, easy, and fast access to machine-learned molecular properties. Compared to its initial version, we present a multitude of new Chemprop functionalities such as the support of multimolecule properties, reactions, atom/bond-level properties, and spectra. Further, we incorporate various uncertainty quantification and calibration methods along with related metrics as well as pretraining and transfer learning workflows, improved hyperparameter optimization, and other customization options concerning loss functions or atom/bond features. We benchmark D-MPNN models trained using Chemprop with the new reaction, atom-level, and spectra functionality on a variety of property prediction data sets, including MoleculeNet and SAMPL, and observe state-of-the-art performance on the prediction of water-octanol partition coefficients, reaction barrier heights, atomic partial charges, and absorption spectra. Chemprop enables out-of-the-box training of D-MPNN models for a variety of problem settings in fast, user-friendly, and open-source software.

Tailoring parameters for QM/MM simulations: accurate modeling of adsorption and catalysis in zirconium-based metal-organic frameworks.

Quantum mechanics/molecular mechanics (QM/MM) simulations offer an efficient way to model reactions occurring in complex environments. This study introduces a specialized set of charge and Lennard-Jones parameters tailored for electrostatically embedded QM/MM calculations, aiming to accurately model both adsorption processes and catalytic reactions in zirconium-based metal-organic frameworks (Zr-MOFs). To validate our approach, we compare adsorption energies derived from QM/MM simulations against experimental results and Monte Carlo simulation outcomes. The developed parameters showcase the ability of QM/MM simulations to represent long-range electrostatic and van der Waals interactions faithfully. This capability is evidenced by the prediction of adsorption energies with a low root mean square error of 1.1 kcal mol-1 across a wide range of adsorbates. The practical applicability of our QM/MM model is further illustrated through the study of glucose isomerization and epimerization reactions catalyzed by two structurally distinct Zr-MOF catalysts, UiO-66 and MOF-808. Our QM/MM calculations closely align with experimental activation energies. Importantly, the parameter set introduced here is compatible with the widely used universal force field (UFF). Moreover, we thoroughly explore how the size of the cluster model and the choice of density functional theory (DFT) methodologies influence the simulation outcomes. This work provides an accurate and computationally efficient framework for modeling complex catalytic reactions within Zr-MOFs, contributing valuable insights into their mechanistic behaviors and facilitating further advancements in this dynamic area of research.

Diels-Alder Conversion of Acrylic Acid and 2,5-Dimethylfuran to para-Xylene Over Heterogeneous Bi-BTC Metal-Organic Framework Catalysts Under Mild Conditions.

The heterogeneous metal-organic framework Bi-BTC successfully catalyzed the synthesis of para-xylene from bio-based 2,5-dimethylfuran and acrylic acid in a promising yield (92 %), under relatively mild conditions (160 °C, 10 bar), and with a low reaction-energy barrier (47.3 kJ mol-1 ). The proposed reaction strategy also demonstrates a remarkable versatility for furan derivatives such as furan and 2-methylfuran.

Effect of mask coverage on face identification in Taiwanese men and women.

Mask wearing is the easiest and most effective way to avoid COVID-19 infection; however, it affects interpersonal activities, especially face identification. This study examined the effects of three mask coverage levels (full coverage, FC; coverage up to the middle [MB] or bottom of the nose bridge [BB]) on face identification accuracy and time. A total of 115 university students (60 men and 55 women) were recruited to conduct a computer-based simulation test consisting of 30 questions (10 questions [five face images each of men and women] for the three mask coverage levels). One unmasked target face and four face images with a specified mask coverage level were designed for each question, and the participants were requested to select the same face from the four covered face images on the basis of the target face. The ANOVA results indicated that identification accuracy was significantly affected by sex (p < 0.01) and the mask coverage level (p < 0.001), whereas identification time was only influenced by sex (p < 0.05). The multiple comparison results indicated that the identification accuracy rate for faces wearing a mask with FC (90.3%) was significantly lower than for those wearing masks with coverage up to the MB (93.7%) and BB (94.9%) positions; however, no difference in identification accuracy rate was observed between the MB and BB levels. Women exhibited a higher identification accuracy rate than men (94.1% vs. 91.9%) in identifying unfamiliar faces, even though they may spend less time identifying the images. A smaller mask coverage level (i.e., the BB level) does not facilitate face identification. The findings can be served as a reference for people to trade-off between wearing a mask and interpersonal interaction in their daily activities.

Autonomous, multiproperty-driven molecular discovery: From predictions to measurements and back.

A closed-loop, autonomous molecular discovery platform driven by integrated machine learning tools was developed to accelerate the design of molecules with desired properties. We demonstrated two case studies on dye-like molecules, targeting absorption wavelength, lipophilicity, and photooxidative stability. In the first study, the platform experimentally realized 294 unreported molecules across three automatic iterations of molecular design-make-test-analyze cycles while exploring the structure-function space of four rarely reported scaffolds. In each iteration, the property prediction models that guided exploration learned the structure-property space of diverse scaffold derivatives, which were realized with multistep syntheses and a variety of reactions. The second study exploited property models trained on the explored chemical space and previously reported molecules to discover nine top-performing molecules within a lightly explored structure-property space.

Comparative Analysis of Uncoupled Mode Approximations for Molecular Thermochemistry and Kinetics.

The accurate prediction of thermochemistry and kinetic parameters is an important task for reaction modeling. Unfortunately, the commonly used harmonic oscillator model is often not accurate enough due to the absence of anharmonic effects. In this work, we improve the representation of an anharmonic potential energy surface (PES) using uncoupled mode (UM) approximations, which model the full-dimensional PES as a sum of one-dimensional potentials of each mode. We systematically analyze different PES sampling schemes and coordinate systems for constructing the one-dimensional potentials, and benchmark the performance of UM methods on data sets of molecular thermochemistry and kinetic properties. The results show that the accuracy of the UM approach strongly depends on how the one-dimensional potentials are defined. If one-dimensional potentials are constructed by sampling along normal mode directions (UM-N) or along the directions that minimize intermode coupling (E- and E'-optimized), the accuracies of the predicted properties are not significantly improved compared to the harmonic oscillator model. However, significant improvements can be achieved by sampling the torsional modes separately from the vibrational modes (UM-T and UM-VT). In this work, three types of coordinate systems are examined, including redundant internal coordinates (RIC), hybrid internal coordinates (HIC), and translation-rotation-internal coordinates (TRIC). The HIC and TRIC coordinate systems can outperform RIC since transition state species may contain large-amplitude interfragmentary motions that regular internal coordinates can not describe adequately. Among all the methods we examined, the activation energies and pre-exponential factors calculated using UM-VT with either TRIC or HIC best agree with the reference values. Since UM-VT requires only a number of additional single point energy calculations for each independent mode, the scaling of computational costs of UM-VT is the same as that of the standard harmonic oscillator model, making UM-VT an appealing way of calculating the thermochemistry and kinetic properties for large-size systems.

Proton-Conductive Cerium-Based Metal-Organic Frameworks.

In this study, proton-conducting behaviors of a cerium-based metal-organic framework (MOF), Ce-MOF-808, its zirconium-based isostructural MOF, and bimetallic MOFs with various Zr-to-Ce ratios are investigated. The significantly increased proton conductivity (σ) and decreased activation energy (Ea) are obtained by substituting Zr with Ce in the nodes of MOF-808. Ce-MOF-808 achieves a σ of 4.4 × 10-3 S/cm at 25 °C under 99% relative humidity and an Ea of 0.14 eV; this value is among the lowest-reported Ea of proton-conductive MOFs. Density functional theory calculations are utilized to probe the proton affinities of these MOFs. As the first study reporting the proton conduction in cerium-based MOFs, the finding here suggests that cerium-based MOFs should be a better platform for the design of proton conductors compared to the commonly reported zirconium-based MOFs in future studies on energy-related applications.

Quantum Mechanical Calculations for Biomass Valorization over Metal-Organic Frameworks (MOFs).

Metal-organic framework (MOF) in biomass valorization is a promising technology developed in recent decades. By tailoring both the metal nodes and organic ligands, MOFs exhibit multiple functionalities, which not only extend their applicability in biomass conversion but also increase the complexity of material designs. To address this issue, quantum mechanical simulations have been used to provide mechanistic insights into the catalysis of biomass-derived molecules, which could potentially facilitate the development of novel MOF-based materials for biomass valorization. The aim of this review is to survey recent quantum mechanical simulations on biomass reactions occurring in MOF catalysts, with the emphasis on the studies of the catalytic activity of active sites and the effects of organic ligand and porous structures on the kinetics. Moreover, different model systems and computational methods used for MOF simulations are also surveyed and discussed in this review.