AQuaRef: Machine learning accelerated quantum refinement of protein structures.

Cryo-EM and X-ray crystallography provide crucial experimental data for obtaining atomic-detail models of biomacromolecules. Refining these models relies on library- based stereochemical restraints, which, in addition to being limited to known chemical entities, do not include meaningful noncovalent interactions relying solely on nonbonded repulsions. Quantum mechanical (QM) calculations could alleviate these issues but are too expensive for large molecules. We present a novel AI-enabled Quantum Refinement (AQuaRef) based on AIMNet2 neural network potential mimicking QM at substantially lower computational costs. By refining 41 cryo-EM and 30 X-ray structures, we show that this approach yields atomic models with superior geometric quality compared to standard techniques, while maintaining an equal or better fit to experimental data.

A community effort in SARS-CoV-2 drug discovery.

The COVID-19 pandemic continues to pose a substantial threat to human lives and is likely to do so for years to come. Despite the availability of vaccines, searching for efficient small-molecule drugs that are widely available, including in low- and middle-income countries, is an ongoing challenge. In this work, we report the results of an open science community effort, the "Billion molecules against COVID-19 challenge", to identify small-molecule inhibitors against SARS-CoV-2 or relevant human receptors. Participating teams used a wide variety of computational methods to screen a minimum of 1 billion virtual molecules against 6 protein targets. Overall, 31 teams participated, and they suggested a total of 639,024 molecules, which were subsequently ranked to find 'consensus compounds'. The organizing team coordinated with various contract research organizations (CROs) and collaborating institutions to synthesize and test 878 compounds for biological activity against proteases (Nsp5, Nsp3, TMPRSS2), nucleocapsid N, RdRP (only the Nsp12 domain), and (alpha) spike protein S. Overall, 27 compounds with weak inhibition/binding were experimentally identified by binding-, cleavage-, and/or viral suppression assays and are presented here. Open science approaches such as the one presented here contribute to the knowledge base of future drug discovery efforts in finding better SARS-CoV-2 treatments.

Synergy of semiempirical models and machine learning in computational chemistry.

Catalyzed by enormous success in the industrial sector, many research programs have been exploring data-driven, machine learning approaches. Performance can be poor when the model is extrapolated to new regions of chemical space, e.g., new bonding types, new many-body interactions. Another important limitation is the spatial locality assumption in model architecture, and this limitation cannot be overcome with larger or more diverse datasets. The outlined challenges are primarily associated with the lack of electronic structure information in surrogate models such as interatomic potentials. Given the fast development of machine learning and computational chemistry methods, we expect some limitations of surrogate models to be addressed in the near future; nevertheless spatial locality assumption will likely remain a limiting factor for their transferability. Here, we suggest focusing on an equally important effort-design of physics-informed models that leverage the domain knowledge and employ machine learning only as a corrective tool. In the context of material science, we will focus on semi-empirical quantum mechanics, using machine learning to predict corrections to the reduced-order Hamiltonian model parameters. The resulting models are broadly applicable, retain the speed of semiempirical chemistry, and frequently achieve accuracy on par with much more expensive ab initio calculations. These early results indicate that future work, in which machine learning and quantum chemistry methods are developed jointly, may provide the best of all worlds for chemistry applications that demand both high accuracy and high numerical efficiency.

Extending machine learning beyond interatomic potentials for predicting molecular properties.

Machine learning (ML) is becoming a method of choice for modelling complex chemical processes and materials. ML provides a surrogate model trained on a reference dataset that can be used to establish a relationship between a molecular structure and its chemical properties. This Review highlights developments in the use of ML to evaluate chemical properties such as partial atomic charges, dipole moments, spin and electron densities, and chemical bonding, as well as to obtain a reduced quantum-mechanical description. We overview several modern neural network architectures, their predictive capabilities, generality and transferability, and illustrate their applicability to various chemical properties. We emphasize that learned molecular representations resemble quantum-mechanical analogues, demonstrating the ability of the models to capture the underlying physics. We also discuss how ML models can describe non-local quantum effects. Finally, we conclude by compiling a list of available ML toolboxes, summarizing the unresolved challenges and presenting an outlook for future development. The observed trends demonstrate that this field is evolving towards physics-based models augmented by ML, which is accompanied by the development of new methods and the rapid growth of user-friendly ML frameworks for chemistry.

Artificial intelligence-enhanced quantum chemical method with broad applicability.

High-level quantum mechanical (QM) calculations are indispensable for accurate explanation of natural phenomena on the atomistic level. Their staggering computational cost, however, poses great limitations, which luckily can be lifted to a great extent by exploiting advances in artificial intelligence (AI). Here we introduce the general-purpose, highly transferable artificial intelligence-quantum mechanical method 1 (AIQM1). It approaches the accuracy of the gold-standard coupled cluster QM method with high computational speed of the approximate low-level semiempirical QM methods for the neutral, closed-shell species in the ground state. AIQM1 can provide accurate ground-state energies for diverse organic compounds as well as geometries for even challenging systems such as large conjugated compounds (fullerene C60) close to experiment. This opens an opportunity to investigate chemical compounds with previously unattainable speed and accuracy as we demonstrate by determining geometries of polyyne molecules-the task difficult for both experiment and theory. Noteworthy, our method's accuracy is also good for ions and excited-state properties, although the neural network part of AIQM1 was never fitted to these properties.

Teaching a neural network to attach and detach electrons from molecules.

Interatomic potentials derived with Machine Learning algorithms such as Deep-Neural Networks (DNNs), achieve the accuracy of high-fidelity quantum mechanical (QM) methods in areas traditionally dominated by empirical force fields and allow performing massive simulations. Most DNN potentials were parametrized for neutral molecules or closed-shell ions due to architectural limitations. In this work, we propose an improved machine learning framework for simulating open-shell anions and cations. We introduce the AIMNet-NSE (Neural Spin Equilibration) architecture, which can predict molecular energies for an arbitrary combination of molecular charge and spin multiplicity with errors of about 2-3 kcal/mol and spin-charges with error errors ~0.01e for small and medium-sized organic molecules, compared to the reference QM simulations. The AIMNet-NSE model allows to fully bypass QM calculations and derive the ionization potential, electron affinity, and conceptual Density Functional Theory quantities like electronegativity, hardness, and condensed Fukui functions. We show that these descriptors, along with learned atomic representations, could be used to model chemical reactivity through an example of regioselectivity in electrophilic aromatic substitution reactions.

Machine learned Hückel theory: Interfacing physics and deep neural networks.

The Hückel Hamiltonian is an incredibly simple tight-binding model known for its ability to capture qualitative physics phenomena arising from electron interactions in molecules and materials. Part of its simplicity arises from using only two types of empirically fit physics-motivated parameters: the first describes the orbital energies on each atom and the second describes electronic interactions and bonding between atoms. By replacing these empirical parameters with machine-learned dynamic values, we vastly increase the accuracy of the extended Hückel model. The dynamic values are generated with a deep neural network, which is trained to reproduce orbital energies and densities derived from density functional theory. The resulting model retains interpretability, while the deep neural network parameterization is smooth and accurate and reproduces insightful features of the original empirical parameterization. Overall, this work shows the promise of utilizing machine learning to formulate simple, accurate, and dynamically parameterized physics models.

Effect of Microenvironment on the Geometrical Structure of d(A)5 d(T)5 and d(G)5 d(C)5 DNA Mini-Helixes and the Dickerson Dodecamer: A Density Functional Theory Study.

We report a comprehensive quantum-chemical study on d(A)5·d(T)5 and d(G)5·d(C)5 DNA mini-helixes and the Dickerson dodecamer d[CGCGAATTCGCG]. The research was performed to model the evolution of the spatial structure of d(A)5·d(T)5 and d(G)5 d(C)5 DNA mini-helixes all the way from vacuum to water bulk. The influence of external factors such as the presence of counterions and the extent of hydration was included. Also, for comparison, limited calculations have been carried out on the Dickerson dodecamer. The study has been performed at the density functional theory level using B97D3 and ωB97XD exchange-correlation functionals augmented by the Def2SVP basis set. We found that the (dA)5·(dT)5 anion when placed in vacuum forms a DNA duplex, which possesses an intermediate form between a helix and a ladder. The presence of compensating Na+ counterions or explicit microhydration of minor and major grooves stabilizes a DNA mini-helix of B-shape. Factors such as water bulk play a minor role. Somewhat different behavior has been found in the case of the (dG)5·(dC)5 duplex. In this case, we observe the formation of B-type mini-helixes even for the (dG)5·(dC)5 anion placed in vacuum. This is due to an additional stabilization originated from the appearance of an extra hydrogen bond, compared to an AT base pair. To assess whether the obtained results are transferable to different sizes of mini-helixes, similar calculations have been performed for the duplex formed by the Dickerson dodecamer which contains a total of 12 dG·dC and dA·dT base pairs. It has been found that in vacuum, analogous to the d(A)5·d(T)5 duplex, this system possesses a shape which is also quite close to a ladder. However, the presence of factors such as hydration restores the B-type geometry. Also, our results completely in line with the results of electrospray-ionization experiments suggest that uncompensated by counterions the DNA backbone preserves the duplex geometry in vacuum. We present arguments that this state is kinetically unstable.

Extending the Applicability of the ANI Deep Learning Molecular Potential to Sulfur and Halogens.

Machine learning (ML) methods have become powerful, predictive tools in a wide range of applications, such as facial recognition and autonomous vehicles. In the sciences, computational chemists and physicists have been using ML for the prediction of physical phenomena, such as atomistic potential energy surfaces and reaction pathways. Transferable ML potentials, such as ANI-1x, have been developed with the goal of accurately simulating organic molecules containing the chemical elements H, C, N, and O. Here, we provide an extension of the ANI-1x model. The new model, dubbed ANI-2x, is trained to three additional chemical elements: S, F, and Cl. Additionally, ANI-2x underwent torsional refinement training to better predict molecular torsion profiles. These new features open a wide range of new applications within organic chemistry and drug development. These seven elements (H, C, N, O, F, Cl, and S) make up ∼90% of drug-like molecules. To show that these additions do not sacrifice accuracy, we have tested this model across a range of organic molecules and applications, including the COMP6 benchmark, dihedral rotations, conformer scoring, and nonbonded interactions. ANI-2x is shown to accurately predict molecular energies compared to density functional theory with a ∼106 factor speedup and a negligible slowdown compared to ANI-1x and shows subchemical accuracy across most of the COMP6 benchmark. The resulting model is a valuable tool for drug development which can potentially replace both quantum calculations and classical force fields for a myriad of applications.

The ANI-1ccx and ANI-1x data sets, coupled-cluster and density functional theory properties for molecules.

Maximum diversification of data is a central theme in building generalized and accurate machine learning (ML) models. In chemistry, ML has been used to develop models for predicting molecular properties, for example quantum mechanics (QM) calculated potential energy surfaces and atomic charge models. The ANI-1x and ANI-1ccx ML-based general-purpose potentials for organic molecules were developed through active learning; an automated data diversification process. Here, we describe the ANI-1x and ANI-1ccx data sets. To demonstrate data diversity, we visualize it with a dimensionality reduction scheme, and contrast against existing data sets. The ANI-1x data set contains multiple QM properties from 5 M density functional theory calculations, while the ANI-1ccx data set contains 500 k data points obtained with an accurate CCSD(T)/CBS extrapolation. Approximately 14 million CPU core-hours were expended to generate this data. Multiple QM calculated properties for the chemical elements C, H, N, and O are provided: energies, atomic forces, multipole moments, atomic charges, etc. We provide this data to the community to aid research and development of ML models for chemistry.

Accurate and transferable multitask prediction of chemical properties with an atoms-in-molecules neural network.

Atomic and molecular properties could be evaluated from the fundamental Schrodinger's equation and therefore represent different modalities of the same quantum phenomena. Here, we present AIMNet, a modular and chemically inspired deep neural network potential. We used AIMNet with multitarget training to learn multiple modalities of the state of the atom in a molecular system. The resulting model shows on several benchmark datasets state-of-the-art accuracy, comparable to the results of orders of magnitude more expensive DFT methods. It can simultaneously predict several atomic and molecular properties without an increase in the computational cost. With AIMNet, we show a new dimension of transferability: the ability to learn new targets using multimodal information from previous training. The model can learn implicit solvation energy (SMD method) using only a fraction of the original training data and an archive median absolute deviation error of 1.1 kcal/mol compared to experimental solvation free energies in the MNSol database.

Approaching coupled cluster accuracy with a general-purpose neural network potential through transfer learning.

Computational modeling of chemical and biological systems at atomic resolution is a crucial tool in the chemist's toolset. The use of computer simulations requires a balance between cost and accuracy: quantum-mechanical methods provide high accuracy but are computationally expensive and scale poorly to large systems, while classical force fields are cheap and scalable, but lack transferability to new systems. Machine learning can be used to achieve the best of both approaches. Here we train a general-purpose neural network potential (ANI-1ccx) that approaches CCSD(T)/CBS accuracy on benchmarks for reaction thermochemistry, isomerization, and drug-like molecular torsions. This is achieved by training a network to DFT data then using transfer learning techniques to retrain on a dataset of gold standard QM calculations (CCSD(T)/CBS) that optimally spans chemical space. The resulting potential is broadly applicable to materials science, biology, and chemistry, and billions of times faster than CCSD(T)/CBS calculations.

Complexation of molecular clips containing fragments of diphenylglycoluril and benzocrown ethers with paraquat and its derivatives.

The complexation of molecular clips containing fragments of diphenylglycoluril and benzocrown ethers with paraquat and its derivatives has been studied both in solution and in the solid state. In this paper we studied the influence of the crown ether ring size and the nature of the substituents at the nitrogen atoms of the paraquat derivatives on the composition and stability of these complexes.

d(A)3d(T)3 and d(G)3d(C)3 B-DNA mini-helixes: the DFT/M06-2x and DFT/B97-D3 comparison of geometrical and energetic characteristics.

We report the comprehensive DFT based comparison of geometrical and energetic parameters of the d(A)3·d(T)3 and d(G)3·d(C)3 nucleic acid mini-helixes performed at B97-D3 and M06-2× levels of theory. We studied the ability of mini-helixes to retain the conformation of B-DNA in the gas phase and under the influence of water bulk, uncompensated charges, and counter-ions. The def2-SV(P) and 6-31G(d,p) basis sets have been used for B97-D3 and M06-2× calculations, correspondently. To estimate basis set superposition error, the recently developed semi-empirical procedure that calls geometrical counterpoise type correction for inter- and intra-molecular basis set superposition error (gcp) has been used in the case of def2-SV(P) basis set. We found that both considered DFT functionals predict very similar results for geometrical ad energetic characteristics. We also found that in contrast to average classical molecular dynamics and data of simple geometrical models, both considered DFT functionals predict the existence of duplex specific geometries. A prediction of interaction energies of d(A)3d(T)3 and d(G)3d(C)3 duplexes accomplished in this study also verifies the applied models and confirms reliability of the new computational gcp technique.

Crystal structure of 4-tert-butyl-2-{2-[N-(3,3-dimethyl-2-oxobut-yl)-N-iso-propyl-carbamo-yl]phen-yl}-1-isopropyl-1H-imidazol-3-ium perchlorate.

In the title salt, C26H40N3O2 (+)·ClO4 (-), the positive charge of the organic cation is delocalized between the two N atoms of the imidazole ring. The C N bond distances are 1.338 (2) and 1.327 (3) Å. The substituents on the benzene ring are rotated almost orthogonal with respect to this ring due to the presence of the bulky isopropyl substituents. The dihedral angle between the benzene and imidazole rings is 75.15 (12)°. Three of the O atoms of the anion are disordered over two sets of sites due to rotation around one of the O-Cl bonds. The ratio of the refined occupancies is 0.591 (14):0.409 (14). In the crystal, the cation and perchlorate anion are bound by an N-H⋯O hydrogen bond. In addition, the cation-anion pairs are linked into layers parallel to (001) by multiple weak C-H⋯O hydrogen bonds.

Chlorido-bis-(1,10-phenanthroline-κ(2) N,N')copper(II) chlorido-(1,10-phen-anthroline-κ(2) N,N')(pyridine-2,6-di-carboxyl-ato-κ(3) O (2),N,O (6))manganate(II) methanol monosolvate.

The title complex, [CuCl(C12H8N2)2][Mn(C7H3NO4)Cl(C12H8N2)]·CH3OH, consists of discrete [CuCl(phen)2](+) cations (phen is 1,10-phenanthroline), [MnCl(pydc)(phen)](-) anions (H2pydc is 2,6-pyridine-2,6-di-carb-oxy-lic acid) and one methanol solvent mol-ecule of crystallization per asymmetric unit. It should be noted, that a solvent-masking procedure as implemented in OLEX2 [Dolomanov et al. (2009). J. Appl. Cryst. 42, 339-341 ▶] was used to remove the electronic contribution from one disordered solvent molecule, presumably methanol. Only the atoms used in the refined model are reported in chemical formula and related values. The Cu(II) ion is five-coordinated by two phenanthroline ligands and one chloride ion in a distorted trigonal-bipyramidal geometry. The dihedral angle between the phen ligands is 65.21 (5)°. The Mn(II) ion is six-coordinated by one Cl atom, two N atoms from a phen ligand, as well one N atom and two O atoms from pydc in a distorted octa-hedral coordination geometry, with cis angles ranging from 72.00 (8) to 122.07 (8)° and trans angles ranging from 143.98 (8) to 163.15 (6)°. In the crystal, C-H⋯O, O-H⋯O and C-H⋯Cl hydrogen bonds, cation-anion π-π inter-actions between the phen ring systems with centroid-centroid distances in the range 3.881 (34)-4.123 (36) Å, as well as cation-cation, anion-anion π-π inter-actions between the phen rings with centroid-centroid distances in the range 3.763 (4)-3.99 (5) Šand pydc rings with centroid-centroid distances 3.52 (5) Šlink the various components.

trans-Chloridobis-(ethane-1,2-di-amine-κ(2) N,N')(thio-cyanato-κN)cobalt(III) diammine-tetra-kis-(thio-cyanato-κN)cromate(III).

The title ionic complex [CoCl(NCS)(C2H8N2)2][Cr(NCS)4(NH3)2], which crystallizes as a non-merohedral twin, is build up of a complex cation [CoCl(NCS)(en)2](+) (en is ethane-1,2-di-amine) and the Reinecke's salt anion [Cr(NCS)4(NH3)2](-) as complex counter-ion. A network of N-H⋯S and N-H⋯Cl hydrogen bonds, as well as short S⋯S contacts [3.538 (2) and 3.489 (3) Å], between the NCS groups of the complex anions link the mol-ecules into a three-dimentional supra-molecular network. Intensity statistic indicated twinning by non-mero-hedry with refined weighs of twin components are 0.5662:0.4338.

Role of supramolecular synthons in the formation of the supramolecular architecture of molecular crystals revisited from an energetic viewpoint.

Analysis of the strengths and directionality of intermolecular interactions in the crystals containing only one type of supramolecular synthon allows the suggestion of a general classification of molecular crystals depending on type of their basic structural motifs. All crystals may be divided on four classes namely (I) crystals with isotropic packing of the building units; (II) columnar crystals where the basic structural motif (BSM) is a chain/column; (III) layered crystals with layers as the BSM; (IV) columnar-layered crystals containing chains/columns as the primary basic structural motif and layers as the secondary BSM. Taking into account the participation of different supramolecular synthons in the formation of different levels of the organization of molecular crystals, they may be considered as basic (responsible for the formation of molecular complexes as building units of crystals), primary, secondary and auxiliary, which are involved in the agglomeration of molecules in primary or secondary basic structural motifs or in the packing of these motifs, respectively. The ranking of supramolecular synthons depends on values of energies of intermolecular interactions and it is individual for each crystal.

Tris(2,2'-bipyridine)-copper(II) penta-cyanido-nitro-soferrate(II) methanol disolvate monohydrate.

The title complex [Cu(C10H8N2)3][Fe(CN)5(NO)]·2CH3OH·H2O, consists of discrete [Cu(bpy)3](2+) cations (bpy is 2,2'-bipyridine), [Fe(CN)5NO](2-) anions and solvent mol-ecules of crystallization (two methanol mol-ecules and one water mol-ecules per asymmetric unit). The Cu(II) ion adopts a distorted octa-hedral environment, coordinated by six N atoms from three bpy ligands. The cation charge is balanced by a nitro-prusside counter-anion, which has a slightly distorted octa-hedral coordination geometry. In the crystal, anions and solvent mol-ecules are involved in O-H⋯N and O-H⋯O hydrogen bonds, which form chains along [100]. The cations are located between these chains.

Influence of deuteration and fluorination on the supramolecular architecture of pyridine N-oxide crystals.

To understand how deuterium and fluorine substituents influence the supramolecular architecture of pyridine N-oxide crystals, the crystal structure of 3-fluoropyridine N-oxide (PNO-3F) was determined and the crystal packing motives of non-deuterated pyridine-N-oxide (PNO), partial-deuterated pyridine-N-oxide (PNO-D) and PNO-3F were analyzed based on ab initio quantum-chemical calculations of the intermolecular interaction energy, using the MP2/6-311G(d,p) method. The appearance of the weak-directing substituents deuterium and fluorine leads to significant changes in the crystal organization of the isotropic packing of PNO molecules.