Choosing the right molecular machine learning potential

Quantum-chemistry simulations based on potential energy surfaces of molecules provide invaluable insight into the physicochemical processes at the atomistic level and yield such important observables as reaction rates and spectra. Machine learning potentials promise to significantly reduce the computational cost and hence enable otherwise unfeasible simulations. However, the surging number of such potentials begs the question of which one to choose or whether we still need to develop yet another one. Here, we address this question by evaluating the performance of popular machine learning potentials in terms of accuracy and computational cost. In addition, we deliver structured information for non-specialists in machine learning to guide them through the maze of acronyms, recognize each potential''s main features, and judge what they could expect from each one.

This article provides a lifeline for those lost in the sea of the molecular machine learning potentials by providing a balanced overview and evaluation of popular potentials.     

Physically inspired deep learning of molecular excitations and photoemission spectra

Modern functional materials consist of large molecular building blocks with significant chemical complexity which limits spectroscopic property prediction with accurate first-principles methods. Consequently, a targeted design of materials with tailored optoelectronic properties by high-throughput screening is bound to fail without efficient methods to predict molecular excited-state properties across chemical space. In this work, we present a deep neural network that predicts charged quasiparticle excitations for large and complex organic molecules with a rich elemental diversity and a size well out of reach of accurate many body perturbation theory calculations. The model exploits the fundamental underlying physics of molecular resonances as eigenvalues of a latent Hamiltonian matrix and is thus able to accurately describe multiple resonances simultaneously. The performance of this model is demonstrated for a range of organic molecules across chemical composition space and configuration space. We further showcase the model capabilities by predicting photoemission spectra at the level of the GW approximation for previously unseen conjugated molecules.

A physically-inspired machine learning model for orbital energies is developed that can be augmented with delta learning to obtain photoemission spectra, ionization potentials, and electron affinities with experimental accuracy.     

Advancing global chemical education through interactive teaching tools

This perspective highlights our recent efforts to develop interactive resources in chemical education for worldwide usage. First, we highlight online tutorials that connect organic chemistry to medicine and popular culture, along with game-like resources for active learning. Next, we describe efforts to aid students in learning to visualize chemical structures in three dimensions. Finally, we present recent approaches toward engaging children and the general population through organic chemistry coloring and activity books. Collectively, these tools have benefited hundreds of thousands of users worldwide. We hope this perspective promotes a spirit of innovation in chemical education and spurs the development of additional free, interactive, and widely accessible chemical education resources.

This perspective highlights the development of interactive chemical education resources for worldwide usage. We hope to promote a spirit of innovation in chemical education and spur the development of new chemical education resources.     

Machine learning of solvent effects on molecular spectra and reactions

Fast and accurate simulation of complex chemical systems in environments such as solutions is a long standing challenge in theoretical chemistry. In recent years, machine learning has extended the boundaries of quantum chemistry by providing highly accurate and efficient surrogate models of electronic structure theory, which previously have been out of reach for conventional approaches. Those models have long been restricted to closed molecular systems without accounting for environmental influences, such as external electric and magnetic fields or solvent effects. Here, we introduce the deep neural network FieldSchNet for modeling the interaction of molecules with arbitrary external fields. FieldSchNet offers access to a wealth of molecular response properties, enabling it to simulate a wide range of molecular spectra, such as infrared, Raman and nuclear magnetic resonance. Beyond that, it is able to describe implicit and explicit molecular environments, operating as a polarizable continuum model for solvation or in a quantum mechanics/molecular mechanics setup. We employ FieldSchNet to study the influence of solvent effects on molecular spectra and a Claisen rearrangement reaction. Based on these results, we use FieldSchNet to design an external environment capable of lowering the activation barrier of the rearrangement reaction significantly, demonstrating promising venues for inverse chemical design.

A machine learning approach for modeling the influence of external environments and fields on molecules has been developed, which allows the prediction of various types of molecular spectra in vacuum and under implicit and explicit solvation.     

DeepReac+: deep active learning for quantitative modeling of organic chemical reactions

Various computational methods have been developed for quantitative modeling of organic chemical reactions; however, the lack of universality as well as the requirement of large amounts of experimental data limit their broad applications. Here, we present DeepReac+, an efficient and universal computational framework for prediction of chemical reaction outcomes and identification of optimal reaction conditions based on deep active learning. Under this framework, DeepReac is designed as a graph-neural-network-based model, which directly takes 2D molecular structures as inputs and automatically adapts to different prediction tasks. In addition, carefully-designed active learning strategies are incorporated to substantially reduce the number of necessary experiments for model training. We demonstrate the universality and high efficiency of DeepReac+ by achieving the state-of-the-art results with a minimum of labeled data on three diverse chemical reaction datasets in several scenarios. Collectively, DeepReac+ has great potential and utility in the development of AI-aided chemical synthesis. DeepReac+ is freely accessible at https://github.com/bm2-lab/DeepReac.

Based on GNNs and active learning, DeepReac+ is designed as a universal framework for quantitative modeling of chemical reactions. It takes molecular structures as inputs directly and adapts to various prediction tasks with fewer training data.     

Making thermodynamic models of mixtures predictive by machine learning: matrix completion of pair interactions

Predictive models of thermodynamic properties of mixtures are paramount in chemical engineering and chemistry. Classical thermodynamic models are successful in generalizing over (continuous) conditions like temperature and concentration. On the other hand, matrix completion methods (MCMs) from machine learning successfully generalize over (discrete) binary systems; these MCMs can make predictions without any data for a given binary system by implicitly learning commonalities across systems. In the present work, we combine the strengths from both worlds in a hybrid approach. The underlying idea is to predict the pair-interaction energies, as they are used in basically all physical models of liquid mixtures, by an MCM. As an example, we embed an MCM into UNIQUAC, a widely-used physical model for the Gibbs excess energy. We train the resulting hybrid model in a Bayesian machine-learning framework on experimental data for activity coefficients in binary systems of 1146 components from the Dortmund Data Bank. We thereby obtain, for the first time, a complete set of UNIQUAC parameters for all binary systems of these components, which allows us to predict, in principle, activity coefficients at arbitrary temperature and composition for any combination of these components, not only for binary but also for multicomponent systems. The hybrid model even outperforms the best available physical model for predicting activity coefficients, the modified UNIFAC (Dortmund) model.

Embedding matrix completion methods from machine learning in classical thermodynamic models creates powerful hybrid models for predicting properties of mixtures.     

An open source computational workflow for the discovery of autocatalytic networks in abiotic reactions

A central question in origins of life research is how non-entailed chemical processes, which simply dissipate chemical energy because they can do so due to immediate reaction kinetics and thermodynamics, enabled the origin of highly-entailed ones, in which concatenated kinetically and thermodynamically favorable processes enhanced some processes over others. Some degree of molecular complexity likely had to be supplied by environmental processes to produce entailed self-replicating processes. The origin of entailment, therefore, must connect to fundamental chemistry that builds molecular complexity. We present here an open-source chemoinformatic workflow to model abiological chemistry to discover such entailment. This pipeline automates generation of chemical reaction networks and their analysis to discover novel compounds and autocatalytic processes. We demonstrate this pipeline''s capabilities against a well-studied model system by vetting it against experimental data. This workflow can enable rapid identification of products of complex chemistries and their underlying synthetic relationships to help identify autocatalysis, and potentially self-organization, in such systems. The algorithms used in this study are open-source and reconfigurable by other user-developed workflows.

We present an open-source chemoinformatic workflow to generate and analyze complex abiological chemical networks to discover novel compounds and autocatalytic processes. We demonstrate this pipeline''s capabilities against a well-studied model system.     

Single-molecule magnets beyond a single lanthanide ion: the art of coupling

The promising future of storing and processing quantized information at the molecular level has been attracting the study of Single-Molecule Magnets (SMMs) for almost three decades. Although some recent breakthroughs are mainly about the SMMs containing only one lanthanide ion, we believe SMMs can tell a much deeper story than the single-ion anisotropy. Here in this Perspective, we will try to draw a unified picture of SMMs as a delicately coupled spin system between multiple spin centres. The hierarchical couplings will be presented step-by-step, from the intra-atomic hyperfine coupling, to the direct and indirect intra-molecular couplings with neighbouring spin centres, and all the way to the inter-molecular and spin–phonon couplings. Along with the discussions on their distinctive impacts on the energy level structures and thus magnetic behaviours, a promising big picture for further studies is proposed, encouraging the multifaceted developments of molecular magnetism and beyond.

In this Perspective, we draw a unified picture for single-molecule magnets as delicately coupled spin systems, discuss the hierarchical couplings (from intra-atomic to inter-molecular) and their distinctive impacts on the magnetic behaviours.     

Machine learning potential era of zeolite simulation

Zeolites, owing to their great variety and complexity in structure and wide applications in chemistry, have long been the hot topic in chemical research. This perspective first presents a short retrospect of theoretical investigations on zeolites using the tools from classical force fields to quantum mechanics calculations and to the latest machine learning (ML) potential simulations. ML potentials as the next-generation technique for atomic simulation open new avenues to simulate and interpret zeolite systems and thus hold great promise for finally predicting the structure–functionality relation of zeolites. Recent advances using ML potentials are then summarized from two main aspects: the origin of zeolite stability and the mechanism of zeolite-related catalytic reactions. We also discussed the possible scenarios of ML potential application aiming to provide instantaneous and easy access of zeolite properties. These advanced applications could now be accomplished by combining cloud-computing-based techniques with ML potential-based atomic simulations. The future development of ML potentials for zeolites in the respects of improving the calculation accuracy, expanding the application scope and constructing the zeolite-related datasets is finally outlooked.

The machine learning atomic simulation will usher the research of zeolite, as other complex materials, into a new era featuring the easy access to zeolite functionalities predicted from theory.     

Diaryliodonium salts facilitate metal-free mechanoredox free radical polymerizations

Mechanically-induced redox processes offer a promising alternative to more conventional thermal and photochemical synthetic methods. For macromolecule synthesis, current methods utilize sensitive transition metal additives and suffer from background reactivity. Alternative methodology will offer exquisite control over these stimuli-induced mechanoredox reactions to couple force with redox-driven chemical transformations. Herein, we present the iodonium-initiated free-radical polymerization of (meth)acrylate monomers under ultrasonic irradiation and ball-milling conditions. We explore the kinetic and structural consequences of these complementary mechanical inputs to access high molecular weight polymers. This methodology will undoubtedly find broad utility across stimuli-controlled polymerization reactions and adaptive material design.

Mechanically-induced redox processes offer a promising alternative to more conventional thermal and photochemical synthetic methods.     

Analog quantum simulation of chemical dynamics

Ultrafast chemical reactions are difficult to simulate because they involve entangled, many-body wavefunctions whose computational complexity grows rapidly with molecular size. In photochemistry, the breakdown of the Born–Oppenheimer approximation further complicates the problem by entangling nuclear and electronic degrees of freedom. Here, we show that analog quantum simulators can efficiently simulate molecular dynamics using commonly available bosonic modes to represent molecular vibrations. Our approach can be implemented in any device with a qudit controllably coupled to bosonic oscillators and with quantum hardware resources that scale linearly with molecular size, and offers significant resource savings compared to digital quantum simulation algorithms. Advantages of our approach include a time resolution orders of magnitude better than ultrafast spectroscopy, the ability to simulate large molecules with limited hardware using a Suzuki–Trotter expansion, and the ability to implement realistic system-bath interactions with only one additional interaction per mode. Our approach can be implemented with current technology; e.g., the conical intersection in pyrazine can be simulated using a single trapped ion. Therefore, we expect our method will enable classically intractable chemical dynamics simulations in the near term.

Dynamics governing ultrafast chemical reactions can be efficiently simulated using analog quantum simulators composed of a coupled system of qudits and bosonic modes.     

Automated discovery of noncovalent inhibitors of SARS-CoV-2 main protease by consensus Deep Docking of 40 billion small molecules

Recent explosive growth of ‘make-on-demand’ chemical libraries brought unprecedented opportunities but also significant challenges to the field of computer-aided drug discovery. To address this expansion of the accessible chemical universe, molecular docking needs to accurately rank billions of chemical structures, calling for the development of automated hit-selecting protocols to minimize human intervention and error. Herein, we report the development of an artificial intelligence-driven virtual screening pipeline that utilizes Deep Docking with Autodock GPU, Glide SP, FRED, ICM and QuickVina2 programs to screen 40 billion molecules against SARS-CoV-2 main protease (Mpro). This campaign returned a significant number of experimentally confirmed inhibitors of Mpro enzyme, and also enabled to benchmark the performance of twenty-eight hit-selecting strategies of various degrees of stringency and automation. These findings provide new starting scaffolds for hit-to-lead optimization campaigns against Mpro and encourage the development of fully automated end-to-end drug discovery protocols integrating machine learning and human expertise.

Deep learning-accelerated docking coupled with computational hit selection strategies enable the identification of inhibitors for the SARS-CoV-2 main protease from a chemical library of 40 billion small molecules.     

Model agnostic generation of counterfactual explanations for molecules

An outstanding challenge in deep learning in chemistry is its lack of interpretability. The inability of explaining why a neural network makes a prediction is a major barrier to deployment of AI models. This not only dissuades chemists from using deep learning predictions, but also has led to neural networks learning spurious correlations that are difficult to notice. Counterfactuals are a category of explanations that provide a rationale behind a model prediction with satisfying properties like providing chemical structure insights. Yet, counterfactuals have been previously limited to specific model architectures or required reinforcement learning as a separate process. In this work, we show a universal model-agnostic approach that can explain any black-box model prediction. We demonstrate this method on random forest models, sequence models, and graph neural networks in both classification and regression.

Generating model agnostic molecular counterfactual explanations to explain model predictions.     

Proton to hydride umpolung at a phosphonium center via electron relay: a new strategy for main-group based water reduction

Generation of dihydrogen from water splitting, also known as water reduction, is a key process to access a sustainable hydrogen economy for energy production and usage. The key step is the selective reduction of a protic hydrogen to an accessible and reactive hydride, which has proven difficult at a p-block element. Although frustrated Lewis pair (FLP) chemistry is well known for water activation by heterolytic H–OH bond cleavage, to the best of our knowledge, there has been only one case showing water reduction by metal-free FLP systems to date, in which silylene (Si^II) was used as the Lewis base. This work reports the molecular design and synthesis of an ortho-phenylene linked bisborane-functionalized phosphine, which reacts with water stoichiometrically to generate H₂ and phosphine oxide quantitatively under ambient conditions. Computational investigations revealed an unprecedented multi-centered electron relay mechanism offered by the molecular framework, shuttling a pair of electrons from hydroxide (OH⁻) in water to the separated proton through a borane-phosphonium-borane path. This simple molecular design and its water reduction mechanism opens new avenues for this main-group chemistry in their growing roles in chemical transformations.

A (bisborane)triarylphosphine was developed to spontaneously generate H₂ from water under ambient conditions, revealing an unprecedented multi-centered electron relay mechanism for a metal-free umpolung of proton to hydride.     

MolE8: finding DFT potential energy surface minima values from force-field optimised organic molecules with new machine learning representations

The use of machine learning techniques in computational chemistry has gained significant momentum since large molecular databases are now readily available. Predictions of molecular properties using machine learning have advantages over the traditional quantum mechanics calculations because they can be cheaper computationally without losing the accuracy. We present a new extrapolatable and explainable molecular representation based on bonds, angles and dihedrals that can be used to train machine learning models. The trained models can accurately predict the electronic energy and the free energy of small organic molecules with atom types C, H N and O, with a mean absolute error of 1.2 kcal mol⁻¹. The models can be extrapolated to larger organic molecules with an average error of less than 3.7 kcal mol⁻¹ for 10 or fewer heavy atoms, which represent a chemical space two orders of magnitude larger. The rapid energy predictions of multiple molecules, up to 7 times faster than previous ML models of similar accuracy, has been achieved by sampling geometries around the potential energy surface minima. Therefore, the input geometries do not have to be located precisely on the minima and we show that accurate density functional theory energy predictions can be made from force-field optimised geometries with a mean absolute error 2.5 kcal mol⁻¹.

New representations and machine learning calculate DFT minima from force field geometries.     

Ultrafast coherent photoexcited dynamics in a trimeric dendrimer probed by X-ray stimulated-Raman signals

The photoinduced ultrafast coherent inter-chromophore energy redistribution in a triarylamine trimer is explored using nonadiabatic excited state molecular dynamics followed by simulations of X-ray Raman signals. The nitrogencentered system ensures strong interchromophore interactions and, thus, the presence of coherences. Nevertheless, the multitude of non-deterministic photoinduced pathways during the ultrafast inter-branch migration of the excitation results in random confinement on some branches and, therefore, spatial exciton scrambling and loss of phase information at long times. We show that the vibronic coherence dynamics evolving into the incoherent scrambling mechanism on ultrafast 50 fs timescale, is accurately probed by the TRUECARS X-ray stimulated Raman signal. In combination with previous results, where the technique has revealed long-lived coherences in a rigid heterodimer, the signal is most valuable for detecting ultrafast molecular coherences or their absence. We demonstrate that X-ray Raman spectroscopy is a useful tool in the chemical design of functional molecular building blocks.

The photoinduced ultrafast coherent inter-chromophore energy redistribution in a triarylamine trimer is explored using nonadiabatic excited state molecular dynamics followed by simulations of X-ray Raman signals.     

Natural product drug discovery in the artificial intelligence era

Natural products (NPs) are primarily recognized as privileged structures to interact with protein drug targets. Their unique characteristics and structural diversity continue to marvel scientists for developing NP-inspired medicines, even though the pharmaceutical industry has largely given up. High-performance computer hardware, extensive storage, accessible software and affordable online education have democratized the use of artificial intelligence (AI) in many sectors and research areas. The last decades have introduced natural language processing and machine learning algorithms, two subfields of AI, to tackle NP drug discovery challenges and open up opportunities. In this article, we review and discuss the rational applications of AI approaches developed to assist in discovering bioactive NPs and capturing the molecular “patterns” of these privileged structures for combinatorial design or target selectivity.

Natural products (NPs) are primarily recognized as privileged structures to interact with protein drug targets.     

An electric field-based approach for quantifying effective volumes and radii of chemically affected space

Chemical shape and size play a critical role in chemistry. The van der Waals (vdW) radius, a familiar manifold used to quantify size by assuming overlapping spheres, provides rapid estimates of size in atoms, molecules, and materials. However, the vdW method may be too rigid to describe highly polarized systems and chemical species that stray from spherical atomistic environments. To deal with these exotic chemistries, numerous alternate methods based on electron density have been presented. While each boasts inherent generality, all define the size of a chemical system, in one way or another, by its electron density. Herein, we revisit the longstanding problem of assessing sizes of atoms and molecules, instead through examination of the local electric field produced by them. While conceptually different than nuclei-centered methods like that of van der Waals, the field assesses chemically affected volumes. This approach implicitly accounts for long-range fields in highly polar systems and predicts that cations should affect more space than neutral counterparts.

Computing atomic and molecular volumes from DFT and ab initio-based electric fields.     

Computational and data driven molecular material design assisted by low scaling quantum mechanics calculations and machine learning

Electronic structure methods based on quantum mechanics (QM) are widely employed in the computational predictions of the molecular properties and optoelectronic properties of molecular materials. The computational costs of these QM methods, ranging from density functional theory (DFT) or time-dependent DFT (TDDFT) to wave-function theory (WFT), usually increase sharply with the system size, causing the curse of dimensionality and hindering the QM calculations for large sized systems such as long polymer oligomers and complex molecular aggregates. In such cases, in recent years low scaling QM methods and machine learning (ML) techniques have been adopted to reduce the computational costs and thus assist computational and data driven molecular material design. In this review, we illustrated low scaling ground-state and excited-state QM approaches and their applications to long oligomers, self-assembled supramolecular complexes, stimuli-responsive materials, mechanically interlocked molecules, and excited state processes in molecular aggregates. Variable electrostatic parameters were also introduced in the modified force fields with the polarization model. On the basis of QM computational or experimental datasets, several ML algorithms, including explainable models, deep learning, and on-line learning methods, have been employed to predict the molecular energies, forces, electronic structure properties, and optical or electrical properties of materials. It can be conceived that low scaling algorithms with periodic boundary conditions are expected to be further applicable to functional materials, perhaps in combination with machine learning to fast predict the lattice energy, crystal structures, and spectroscopic properties of periodic functional materials.

Low scaling quantum mechanics calculations and machine learning can be employed to efficiently predict the molecular energies, forces, and optical and electrical properties of molecular materials and their aggregates.     

Defining the conditions for the development of the emerging class of FeIII-based MRI contrast agents

Fe(iii) complexes are attracting growing interest in chemists developing diagnostic probes for Magnetic Resonance Imaging because they leverage on an endogenous metal and show superior stability. However, in this case a detailed understanding of the relationship between the chemical structure of the complexes, their magnetic, thermodynamic, kinetic and redox properties and the molecular parameters governing the efficacy (relaxivity) is still far from being available. We have carried out an integrated ¹H and ¹⁷O NMR relaxometric study as a function of temperature and magnetic field, on the aqua ion and three complexes chosen as reference models, together with theoretical calculations, to obtain accurate values of the parameters that control their relaxivity. Moreover, thermodynamic stability and dissociation kinetics of the Fe(iii) chelates, measured in association with the ascorbate reduction behaviour, highlight their role and mutual influence in achieving the stability required for use in vivo.

An integrated ¹H and ¹⁷O NMR relaxometric study on model systems allowed to highlight that the Fe(III) complexes might represent the best alternative to Gd-based MRI contrast agents at the magnetic fields of current and future clinical scanners.