Heterocycles as supramolecular handles for crystal engineering: a case study with 7-(diethylamino)coumarin derivatives

In this study, we present the synthesis and detailed solid-state structural characterization of a Schiff-base-bridged derivative of 7-(diethylamino)coumarin (7-DAC), a molecular block displaying repetitive aggregation modes in the solid state despite being attached to broadly different molecular frameworks. To map the supramolecular habits of this unconventional moiety, we carry out a comparative analysis of the crystal packing in a curated dataset of 50 molecules decorated with the 7-DAC group, retrieved from the literature. We uncover that self-recognition of the 7-DAC moiety has two main components: a set of directional C-H⋯O interactions between neighboring coumarins, and antiparallel dipole-dipole interactions, taking the form of distinct π-stacking modes. The pendant 7-diethylamino group is key to the behavior of 7-DAC, favoring its solubilization through its conformational flexibility in solution, while in the crystalline matrix, it acts as a structural spacer that favors π-stacking interactions. Our findings present a comprehensive analysis of the preferential arrangements of the 7-DAC fragment in various (supra)molecular scenarios, confirming that it is (i) a mobile but mostly planar group, (ii) a group prone to antiparallel aggregation, and (iii) up to 90% likely to pack via π-stacking supported by hydrogen-bonding interactions. These findings enrich the palette of supramolecular motifs available for the bottom-up design of organic materials and their programmed construction.

The Relationship Between Molecular Symmetry and Physicochemical Properties Involving Boiling and Melting of Organic Compounds

OBJECTIVE AND METHODS

The reliable estimation of phase transition physicochemical properties such as boiling and melting points can be valuable when designing compounds with desired physicochemical properties. This study explores the role of external rotational symmetry in determining boiling and melting points of select organic compounds. Using experimental data from the literature, the entropies of boiling and fusion were obtained for 541 compounds. The statistical significance of external rotational symmetry number on entropies of phase change was determined by using multiple linear regression. In addition, a series of aliphatic hydrocarbons, polysubstituted benzenes, and di-substituted napthalenes are used as examples to demonstrate the role of external symmetry on transition temperature.

RESULTS

The results reveal that symmetry is not well correlated with boiling point but is statistically significant in melting point.

CONCLUSION

The lack of correlation between the boiling point and the symmetry number reflects the fact that molecules have a high degree of rotational freedom in both the liquid and the vapor. On the other hand, the strong relationship between symmetry and melting point reflects the fact that molecules are rotationally restricted in the crystal but not in the liquid. Since the symmetry number is equal to the number of ways that the molecule can be properly oriented for incorporation into the crystal lattice, it is a significant determinant of the melting point.

General Graph Neural Network-Based Model To Accurately Predict Cocrystal Density and Insight from Data Quality and Feature Representation

Cocrystal engineering as an effective way to modify solid-state properties has inspired great interest from diverse material fields while cocrystal density is an important property closely correlated with the material function. In order to accurately predict the cocrystal density, we develop a graph neural network (GNN)-based deep learning framework by considering three key factors of machine learning (data quality, feature presentation, and model architecture). The result shows that different stoichiometric ratios of molecules in cocrystals can significantly influence the prediction performances, highlighting the importance of data quality. In addition, the feature complementary is not suitable for augmenting the molecular graph representation in the cocrystal density prediction, suggesting that the complementary strategy needs to consider whether extra features can sufficiently supplement the lacked information in the original representation. Based on these results, 4144 cocrystals with 1:1 stoichiometry ratio are selected as the dataset, supplemented by the data augmentation of exchanging a pair of coformers. The molecular graph is determined to learn feature representation to train the GNN-based model. Global attention is introduced to further optimize the feature space and identify important atoms to realize the interpretability of the model. Benefited from the advantages, our model significantly outperforms three competitive models and exhibits high prediction accuracy for unseen cocrystals, showcasing its robustness and generality. Overall, our work not only provides a general cocrystal density prediction tool for experimental investigations but also provides useful guidelines for the machine learning application. All source codes are freely available at https://github.com/Xiao-Gua00/CCPGraph.

Application of gas chromatographic data and 2D molecular descriptors for accurate global mobility potential prediction

Mobility is a key feature affecting the environmental fate, which is of particular importance in the case of persistent organic pollutants (POPs) and emerging pollutants (EPs). In this study, the global mobility classification artificial neural networks-based models employing GC retention times (RT) and 2D molecular descriptors were constructed and validated. The high usability of RT was confirmed based on the feature selection step performed using the multivariate adaptive regression splines (MARS) tool. Although RT was found to be the most important, according to Kruskal-Wallis ANOVA analysis, it is insufficient to build a robust model, which justifies the need to expand the input layer with 2D descriptors. Therefore the following molecular descriptors: MPC10, WTPT-2, AATS8s, minaaCH, GATS7c, RotBtFrac, ATSC7v and ATSC1p, which were characterized by a high predicting potential were used to improve the classification performance. As a result of machine learning procedure ten of the most accurate neural networks were selected. The external validation showed that the final models are characterized by a high general accuracy score (85.71-96.43%). The high predicting abilities were also confirmed by the micro-averaged Matthews correlation coefficient (MAMCC) (0.73-0.88). To evaluate the applicability of the models, new retention times of selected POPs and EPs including pesticides, polycyclic aromatic hydrocarbons, pharmaceuticals, fragrances and personal care products were measured and used for mobility prediction. Further, the classifiers were used for photodegradation and chlorination products of two popular sunscreen agents, 2-ethyl-hexyl-4-methoxycinnamate and 2-ethylhexyl 4-(dimethylamino)benzoate.

Applications of Artificial Intelligence and Machine Learning Algorithms to Crystallization

Artificial intelligence and specifically machine learning applications are nowadays used in a variety of scientific applications and cutting-edge technologies, where they have a transformative impact. Such an assembly of statistical and linear algebra methods making use of large data sets is becoming more and more integrated into chemistry and crystallization research workflows. This review aims to present, for the first time, a holistic overview of machine learning and cheminformatics applications as a novel, powerful means to accelerate the discovery of new crystal structures, predict key properties of organic crystalline materials, simulate, understand, and control the dynamics of complex crystallization process systems, as well as contribute to high throughput automation of chemical process development involving crystalline materials. We critically review the advances in these new, rapidly emerging research areas, raising awareness in issues such as the bridging of machine learning models with first-principles mechanistic models, data set size, structure, and quality, as well as the selection of appropriate descriptors. At the same time, we propose future research at the interface of applied mathematics, chemistry, and crystallography. Overall, this review aims to increase the adoption of such methods and tools by chemists and scientists across industry and academia.

Prediction of Grain Size in Cast Aluminum Alloys

Grain refinement of cast alloys, especially aluminum–silicon and magnesium-based alloys, is an effective approach to improve the strength of alloys. Grain size is the most representative parameter used to characterize grain refinement in the industry, thereby attracting increasing attention for developing accurate grain size prediction models. In this paper, several important grain size prediction models under different adaptation conditions are reviewed. These models are obtained either by regression of experimental data or by physical/mathematical inference under certain assumptions of specified cases, focusing on the effects of alloy composition, solidification temperature gradient, grain growth rate, and fining agent composition, among others. The trends of grain size prediction models were also discussed. The results revealed machine learning as an effective tool to establish a data-driven prediction model of grain size in cast aluminum alloys.

Machine Learning-Driven, Sensor-Integrated Microfluidic Device for Monitoring and Control of Supersaturation for Automated Screening of Crystalline Materials

Integrating sensors in miniaturized devices allow for fast and sensitive detection and precise control of experimental conditions. One of the potential applications of a sensor-integrated microfluidic system is to measure the solute concentration during crystallization. In this study, a continuous-flow microfluidic mixer is paired with an electrochemical sensor to enable in situ measurement of the supersaturation. This sensor is investigated as the predictive measurement of the supersaturation during the antisolvent crystallization of l-histidine in the water-ethanol mixture. Among the various metals tested in a batch system for their sensitivity toward l-histidine, Pt showed the highest sensitivity. A Pt-printed electrode was inserted in the continuous-flow microfluidic mixer, and the cyclic voltammograms of the system were obtained for different concentrations of l-histidine and different water-to-ethanol ratios. The sensor was calibrated for different ratios of antisolvent and concentrations of l-histidine with respect to the change of the measured anodic slope. Additionally, a machine-learning algorithm using neural networks was developed to predict the supersaturation of l-histidine from the measured anodic slope. The electrochemical sensors have shown sensitivity toward l-histidine, l-glutamic acid, and o-aminobenzoic acid, which consist of functional groups present in almost 80% of small-molecule drugs on the market. The machine learning-guided electrochemical sensors can be applied to other small molecules with similar functional groups for automated screening of crystallization conditions in microfluidic devices.

Ligand binding remodels protein side-chain conformational heterogeneity

While protein conformational heterogeneity plays an important role in many aspects of biological function, including ligand binding, its impact has been difficult to quantify. Macromolecular X-ray diffraction is commonly interpreted with a static structure, but it can provide information on both the anharmonic and harmonic contributions to conformational heterogeneity. Here, through multiconformer modeling of time- and space-averaged electron density, we measure conformational heterogeneity of 743 stringently matched pairs of crystallographic datasets that reflect unbound/apo and ligand-bound/holo states. When comparing the conformational heterogeneity of side chains, we observe that when binding site residues become more rigid upon ligand binding, distant residues tend to become more flexible, especially in non-solvent-exposed regions. Among ligand properties, we observe increased protein flexibility as the number of hydrogen bonds decreases and relative hydrophobicity increases. Across a series of 13 inhibitor-bound structures of CDK2, we find that conformational heterogeneity is correlated with inhibitor features and identify how conformational changes propagate differences in conformational heterogeneity away from the binding site. Collectively, our findings agree with models emerging from nuclear magnetic resonance studies suggesting that residual side-chain entropy can modulate affinity and point to the need to integrate both static conformational changes and conformational heterogeneity in models of ligand binding.

In Silico Study Approach on a Series of 50 Polyphenolic Compounds in Plants; A Comparison on the Bioavailability and Bioactivity Data

Fifty (50) phytocompounds from several subclasses of polyphenols, chosen based on their abundance in the plant world, were analyzed through density functional methods, using computational tools to evaluate their oral availability and particular bioactivity on several cell modulators; key descriptors and molecular features related to the electron density and electrostatic potential for the lowest energy conformers of the investigated molecules were computed. An analysis of the bioactivity scores towards six cell modulators (GPCR ligand, ion channel modulator, kinase inhibitor, nuclear receptor ligand, protease inhibitor and enzyme inhibitor) was also achieved, in the context of investigating their potential side effects on the human digestive processes. Summarizing, computational results confirmed in vivo and in vitro data regarding the high bioavailability of soy isoflavones and better bioavailability of free aglycones in comparison with their esterified and glycosylated forms. However, by a computational approach analyzing Lipinski’s rule, apigenin and apigenin-7-O-rhamnoside, naringenin, hesperetin, genistein, daidzin, biochanin A and formonetin in the flavonoid series and all hydroxycinnamic acids and all hydroxybenzoic acids excepting ellagic acid were proved to have the best bioavailability data; rhamnoside derivatives, the predominant glycosides in green plants, which were reported to have the lowest bioavailability values by in vivo studies, were revealed to have the best bioavailability data among the studied flavonoids in the computational approach. Results of in silico screening on the phenolic derivatives series also revealed their real inhibitory potency on the six parameters studied, showing a remarkable similitude between the flavonoid series, while flavonoids were more powerful natural cell modulators than the phenyl carboxylic acids tested. Thus, it can be concluded that there is a need for supplementation with digestive enzymes, mainly in the case of individuals with low digestive efficiency, to obtain the best health benefits of polyphenols in humans.

Progress in Research on Artificial Intelligence Applied to Polymorphism and Cocrystal Prediction

Artificial intelligence (AI) is a technology that builds an artificial system with certain intelligence and uses computer software and hardware to simulate intelligent human behavior. When combined with drug research and development, AI can considerably shorten this cycle, improve research efficiency, and minimize costs. The use of machine learning to discover novel materials and predict material properties has become a new research direction. On the basis of the current status of worldwide research on the combination of AI and crystal form and cocrystal, this mini-review analyzes and explores the application of AI in polymorphism prediction, crystal structure analysis, crystal property prediction, cocrystal former (CCF) screening, cocrystal composition prediction, and cocrystal formation prediction. This study provides insights into the future applications of AI in related fields.

Hydrogen-bond directionality and symmetry in [C(O)NH](N)2P(O)-based structures: a comparison between X-ray crystallography data and neutron-normalized values, and evaluation of reliability

For [C(O)NH](N)₂P(O)-based structures, the magnitude of the differences in the N-H...O, H...O=P and H...O=C angles has been evaluated when the N-H bond lengths, determined by X-ray diffraction, were compared to the neutron normalized values and the maximum percentage difference was obtained, i.e. about 3% for the angle even if the N-H bond lengths have a difference of about 30% (0.7 Å for the X-ray and 1.03 Å for the neutron-normalized value). The symmetries of the crystals are discussed with respect to the symmetry of the molecules, as well as to the symmetry of hydrogen-bonded motifs, and the role of the most directional hydrogen bond in raising the probability of obtaining centrosymmetric crystal structures is investigated. The work was performed by considering nine new X-ray crystal structures and 204 analogous structures retrieved from the Cambridge Structural Database.

Molecular, Solid-State and Surface Structures of the Conformational Polymorphic Forms of Ritonavir in Relation to their Physicochemical Properties

Purpose

Application of multi-scale modelling workflows to characterise polymorphism in ritonavir with regard to its stability, bioavailability and processing.

Methods

Molecular conformation, polarizability and stability are examined using quantum mechanics (QM). Intermolecular synthons, hydrogen bonding, crystal morphology and surface chemistry are modelled using empirical force fields.

Results

The form I conformation is more stable and polarized with more efficient intermolecular packing, lower void space and higher density, however its shielded hydroxyl is only a hydrogen bond donor. In contrast, the hydroxyl in the more open but less stable and polarized form II conformation is both a donor and acceptor resulting in stronger hydrogen bonding and a more stable crystal structure but one that is less dense. Both forms have strong 1D networks of hydrogen bonds and the differences in packing energies are partially offset in form II by its conformational deformation energy difference with respect to form I. The lattice energies converge at shorter distances for form I, consistent with its preferential crystallization at high supersaturation. Both forms exhibit a needle/lath-like crystal habit with slower growing hydrophobic side and faster growing hydrophilic capping habit faces with aspect ratios increasing from polar-protic, polar-aprotic and non-polar solvents, respectively. Surface energies are higher for form II than form I and increase with solvent polarity. The higher deformation, lattice and surface energies of form II are consistent with its lower solubility and hence bioavailability.

Conclusion

Inter-relationship between molecular, solid-state and surface structures of the polymorphic forms of ritonavir are quantified in relation to their physical-chemical properties.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11095-021-03048-2.

Thermal boundary resistance at graphene-pentacene interface explored by a data-intensive approach

As the machinery of artificial intelligence matures in recent years, there has been a surge in applying machine learning (ML) techniques for material property predictions. Artificial neural network (ANN) is a branch of ML and has gained increasing popularity due to its capabilities of modeling complex correlations among large datasets. The interfacial thermal transport plays a significant role in the thermal management of graphene-pentacene based organic electronics. In this work, the thermal boundary resistance (TBR) between graphene and pentacene is comprehensively investigated by classical molecular dynamics simulations combined with the ML technique. The TBR values along thea,bandcdirections of pentacene at 300 K are 5.19 ± 0.18 × 10^-8m²K W^-1, 3.66 ± 0.36 × 10^-8m²K W^-1and 5.03 ± 0.14 × 10^-8m²K W^-1, respectively. Different architectures of ANN models are trained to predict the TBR between graphene and pentacene. Two important hyperparameters, i.e. network layer and the number of neurons are explored to achieve the best prediction results. It is reported that the two-layer ANN with 40 neurons each layer provides the optimal model performance with a normalized mean square error loss of 7.04 × 10^-4. Our results provide reasonable guidelines for the thermal design and development of graphene-pentacene electronic devices.

First global analysis of the GSK database of small molecule crystal structures

Analysis of the molecular and structural features of the GSK crystal structure database and Cambridge Structural Database leads to improved reliability in hydrogen bond propensity models for pharmaceutical polymorphs.

One class classification as a practical approach for accelerating π-π co-crystal discovery

The implementation of machine learning models has brought major changes in the decision-making process for materials design. One matter of concern for the data-driven approaches is the lack of negative data from unsuccessful synthetic attempts, which might generate inherently imbalanced datasets. We propose the application of the one-class classification methodology as an effective tool for tackling these limitations on the materials design problems. This is a concept of learning based only on a well-defined class without counter examples. An extensive study on the different one-class classification algorithms is performed until the most appropriate workflow is identified for guiding the discovery of emerging materials belonging to a relatively small class, that being the weakly bound polyaromatic hydrocarbon co-crystals. The two-step approach presented in this study first trains the model using all the known molecular combinations that form this class of co-crystals extracted from the Cambridge Structural Database (1722 molecular combinations), followed by scoring possible yet unknown pairs from the ZINC15 database (21 736 possible molecular combinations). Focusing on the highest-ranking pairs predicted to have higher probability of forming co-crystals, materials discovery can be accelerated by reducing the vast molecular space and directing the synthetic efforts of chemists. Further on, using interpretability techniques a more detailed understanding of the molecular properties causing co-crystallization is sought after. The applicability of the current methodology is demonstrated with the discovery of two novel co-crystals, namely pyrene-6H-benzo[c]chromen-6-one (1) and pyrene-9,10-dicyanoanthracene (2).

Machine learning using one class classification on a database of existing co-crystals enables the identification of co-formers which are likely to form stable co-crystals, resulting in the synthesis of two co-crystals of polyaromatic hydrocarbons.

Can molecular flexibility control crystallization? The case of para substituted benzoic acids

Despite the technological importance of crystallization from solutions almost nothing is known about the relationship between the kinetic process of nucleation and the molecular and crystal structures of a crystallizing solute. Nowhere is this more apparent than in our attempts to understand the behavior of increasingly large, flexible molecules developed as active components in the pharmaceutical arena. In our current contribution we develop a general protocol involving a combination of computation (conformation analysis, lattice energy), and experiment (measurement of nucleation rates), and show how significant advances can be made. We present the first systematic study aimed at quantifying the impact of molecular flexibility on nucleation kinetics. The nucleation rates of 4 para substituted benzoic acids are compared, two of which have substituents with flexible chains. In making this comparison, the importance of normalizing data to account for differing solubilities is highlighted. These data have allowed us to go beyond popular qualitative descriptors such ‘crystallizability’ or ‘crystallization propensity’ in favour of more precise nucleation rate data. Overall, this leads to definite conclusions as to the relative importance of solution chemistry, solid-state interactions and conformational flexibility in the crystallization of these molecules and confirms the key role of intermolecular stacking interactions in determining relative nucleation rates. In a more general sense, conclusions are drawn as to conditions under which conformational change may become rate determining during a crystallization process.

Little is known about the relationship between the kinetic process of nucleation and the molecular and crystal structures of a crystallizing solute. Here we compare the behaviour of a series of benzoic acids with a focus on conformational effects.

Fundamental aspects of DMPK optimization of targeted protein degraders

Targeted protein degraders are an emerging modality. Their properties fall outside the traditional small-molecule property space and are in the 'beyond rule of 5' space. Consequently, drug discovery programs focused on developing orally bioavailable degraders are expected to face complex drug metabolism and pharmacokinetics (DMPK) challenges compared with traditional small molecules. Nevertheless, little information is available on the DMPK optimization of oral degraders. Therefore, in this review, we discuss our experience of these DMPK optimization challenges and present methodologies and strategies to overcome the hurdles dealing with this new small-molecule modality, specifically in developing oral degraders to treat cancer.

A Short Review of Current Computational Concepts for High-Pressure Phase Transition Studies in Molecular Crystals

High-pressure chemistry of organic compounds is a hot topic of modern chemistry. In this work, basic computational concepts for high-pressure phase transition studies in molecular crystals are described, showing their advantages and disadvantages. The interconnection of experimental and computational methods is highlighted, showing the importance of energy calculations in this field. Based on our deep understanding of methods’ limitations, we suggested the most convenient scheme for the computational study of high-pressure crystal structure changes. Finally, challenges and possible ways for progress in high-pressure phase transitions research of organic compounds are briefly discussed.

Machine learning methods to predict the crystallization propensity of small organic molecules

Machine learning algorithms were explored for the prediction of the crystallization propensity based on molecular descriptors and fingerprints generated from 2D chemical structures and 3D chemical structures optimized with empirical methods.

An experimental and computational study into the crystallisation propensity of 2nd generation sulflower

The crystallisation propensity of the newly synthesised molecule persulfurated coronene has been investigated through a number of experimental methods. Electrostatic potential calculations and multi-molecular optimisations show that face-face interactions are far more favorable than edge-face interactions, severely restricting the ability of the molecule to crystallise.

A Million Crystal Structures: The Whole Is Greater than the Sum of Its Parts

The founding in 1965 of what is now called the Cambridge Structural Database (CSD) has reaped dividends in numerous and diverse areas of chemical research. Each of the million or so crystal structures in the database was solved for its own particular reason, but collected together, the structures can be reused to address a multitude of new problems. In this Review, which is focused mainly on the last 10 years, we chronicle the contribution of the CSD to research into molecular geometries, molecular interactions, and molecular assemblies and demonstrate its value in the design of biologically active molecules and the solid forms in which they are delivered. Its potential in other commercially relevant areas is described, including gas storage and delivery, thin films, and (opto)electronics. The CSD also aids the solution of new crystal structures. Because no scientific instrument is without shortcomings, the limitations of CSD research are assessed. We emphasize the importance of maintaining database quality: notwithstanding the arrival of big data and machine learning, it remains perilous to ignore the principle of garbage in, garbage out. Finally, we explain why the CSD must evolve with the world around it to ensure it remains fit for purpose in the years ahead.

Assessment of machine learning approaches for predicting the crystallization propensity of active pharmaceutical ingredients

This work critically evaluates a number of machine learning approaches for predicting the crystallization propensity of active pharmaceutical ingredients using a real-world dataset.

Machine learning for molecular and materials science

Here we summarize recent progress in machine learning for the chemical sciences. We outline machine-learning techniques that are suitable for addressing research questions in this domain, as well as future directions for the field. We envisage a future in which the design, synthesis, characterization and application of molecules and materials is accelerated by artificial intelligence.

Salt formation improved the properties of a candidate drug during early formulation development

The purpose of this study was to investigate if AZD5329, a dual neurokinin NK1/2 receptor antagonist, is a suitable candidate for further development as an oral immediate release (IR) solid dosage form as a final product. The neutral form of AZD5329 has only been isolated as amorphous material. In order to search for a solid material with improved physical and chemical stability and more suitable solid-state properties, a salt screen was performed. Crystalline material of a maleic acid salt and a fumaric acid salt of AZD5329 were obtained. X-ray powder diffractiometry, thermogravimetric analysis, differential scanning calorimetry and dynamic vapor sorption were used to investigate the physicochemical characteristics of the two salts. The fumarate salt of AZD5329 is anhydrous, the crystallization is reproducible and the hygroscopicity is acceptable. Early polymorphism assessment work using slurry technique did not reveal any better crystal modification or crystallinity for the fumarate salt. For the maleate salt, the form isolated originally was found to be a solvate, but an anhydrous form was found in later experiments; by suspension in water or acetone, by drying of the solvate to 100-120 °C or by subjecting the solvate form to conditions of 40 °C/75%RH for 3 months. The dissolution behavior and the chemical stability (in aqueous solutions, formulations and solid-state) of both salts were also studied and found to be satisfactory. The compound displays sensitivity to low pH, and the salt of the maleic acid, which is the stronger acid, shows more degradation during stability studies, in line with this observation. The presented data indicate that the substance fulfils basic requirements for further development of an IR dosage form, based on the characterization on crystalline salts of AZD5329.

A random forest model for predicting crystal packing of olanzapine solvates

A random forest (RF) classification model obtained from physicochemical properties of solvents and crystal structures of olanzapine has for the first time enabled the prediction of 3-D crystal packings of solvates. A novel solvate was obtained by targeted crystallization from the solvent identified by RF model.

How many ritonavir cases are there still out there?

The number of dormant ritonavir cases is estimated based on 41 commercial pharmaceutical crystal structure prediction studies.

Identifying models of dielectric breakdown strength from high-throughput data via genetic programming

The identification of models capable of rapidly predicting material properties enables rapid screening of large numbers of materials and facilitates the design of new materials. One of the leading challenges for computational researchers is determining the best ways to analyze large material data sets to identify models that can rapidly predict a given property. In this paper, we demonstrate the use of genetic programming to generate simple models of dielectric breakdown based on 82 representative dielectric materials. We identified the band gap E_g and phonon cut-off frequency ω_max as the two most relevant features, and new classes of models featuring functions of E_g and ω_max were uncovered. The genetic programming approach was found to outperform other approaches for generating models, and we discuss some of the advantages of this approach.

Human versus Robots in the Discovery and Crystallization of Gigantic Polyoxometalates

The discovery of new gigantic molecules formed by self-assembly and crystal growth is challenging as it combines two contingent events; first is the formation of a new molecule, and second its crystallization. Herein, we construct a workflow that can be followed manually or by a robot to probe the envelope of both events and employ it for a new polyoxometalate cluster, Na₆ [Mo₁₂₀ Ce₆ O₃₆₆ H₁₂ (H₂ O)₇₈ ]⋅200 H₂ O (1) which has a trigonal-ring type architecture (yield 4.3 % based on Mo). Its synthesis and crystallization was probed using an active machine-learning algorithm developed by us to explore the crystallization space, the algorithm results were compared with those obtained by human experimenters. The algorithm-based search is able to cover ca. 9 times more crystallization space than a random search and ca. 6 times more than humans and increases the crystallization prediction accuracy to 82.4±0.7 % over 77.1±0.9 % from human experimenters.

Will they co-crystallize?

Co-crystal screening data and machine learning models allows prediction of the most likely co-formers to use for new molecules.

A publicly available crystallisation data set and its application in machine learning

A publicly available crystallisation database for clusters of highly similar compounds is used to build machine learning models.

Beyond Rotatable Bond Counts: Capturing 3D Conformational Flexibility in a Single Descriptor

A new molecular descriptor, nConf20, based on chemical connectivity, is presented which captures the accessible conformational space of a molecule. Currently the best available two-dimensional descriptors for quantifying the flexibility of a particular molecule are the rotatable bond count (RBC) and the Kier flexibility index. We present a descriptor which captures this information by sampling the conformational space of a molecule using the RDKit conformer generator. Flexibility has previously been identified as a key feature in determining whether a molecule is likely to crystallize or not. For this application, nConf20 significantly outperforms previously reported single-variable classifiers and also assists rule-based analysis of black-box machine learning classification algorithms.

Computational Screening of Drug Solvates

PURPOSE

Solvates are mainly undesired by-products during the pharmaceutical development of new drugs. In addition, solvate formation may also distort solubility measurements. The presented study introduces a simple computational approach that allows for the identification of drug solvent pairs which most likely form crystalline solid phases.

METHODS

The mixing enthalpy as a measure for drug-solvent complementarity is obtained by computational liquid phase thermodynamics (COSMO-RS theory). In addition a few other simple descriptors were taking into account describing the shape and topology of the drug and the solvent. Using an extensive dataset of drug solvent pairs a simple and statistically robust model is developed which allows for a rough assessment of a solvent's ability to form a solvate.

RESULTS

Similar to the related issue of cocrystal screening, the mixing (or excess) enthalpy of the subcooled liquid mixture of the drug-solvent pair proves to be an important quantity controlling solvate formation. Due to the fact that many solvates form inclusion compounds, the solvent shape is another important factor influencing solvate formation. Solvates forming channel-like voids in the solid state are predicted less well.

CONCLUSION

The approach ranks any drug-solvent pair that forms a solvate before any non-solvate by a probability of about 81% (AUC = 0.81), giving a significant advantage over any trial and error approach. Hence it can help to identify suitable solvent candidates early in the drug development process.

Can computed crystal energy landscapes help understand pharmaceutical solids?

Computational crystal structure prediction (CSP) methods can now be applied to the smaller pharmaceutical molecules currently in drug development. We review the recent uses of computed crystal energy landscapes for pharmaceuticals, concentrating on examples where they have been used in collaboration with industrial-style experimental solid form screening. There is a strong complementarity in aiding experiment to find and characterise practically important solid forms and understanding the nature of the solid form landscape.

Machine-learning-assisted materials discovery using failed experiments

Inorganic-organic hybrid materials such as organically templated metal oxides, metal-organic frameworks (MOFs) and organohalide perovskites have been studied for decades, and hydrothermal and (non-aqueous) solvothermal syntheses have produced thousands of new materials that collectively contain nearly all the metals in the periodic table. Nevertheless, the formation of these compounds is not fully understood, and development of new compounds relies primarily on exploratory syntheses. Simulation- and data-driven approaches (promoted by efforts such as the Materials Genome Initiative) provide an alternative to experimental trial-and-error. Three major strategies are: simulation-based predictions of physical properties (for example, charge mobility, photovoltaic properties, gas adsorption capacity or lithium-ion intercalation) to identify promising target candidates for synthetic efforts; determination of the structure-property relationship from large bodies of experimental data, enabled by integration with high-throughput synthesis and measurement tools; and clustering on the basis of similar crystallographic structure (for example, zeolite structure classification or gas adsorption properties). Here we demonstrate an alternative approach that uses machine-learning algorithms trained on reaction data to predict reaction outcomes for the crystallization of templated vanadium selenites. We used information on 'dark' reactions--failed or unsuccessful hydrothermal syntheses--collected from archived laboratory notebooks from our laboratory, and added physicochemical property descriptions to the raw notebook information using cheminformatics techniques. We used the resulting data to train a machine-learning model to predict reaction success. When carrying out hydrothermal synthesis experiments using previously untested, commercially available organic building blocks, our machine-learning model outperformed traditional human strategies, and successfully predicted conditions for new organically templated inorganic product formation with a success rate of 89 per cent. Inverting the machine-learning model reveals new hypotheses regarding the conditions for successful product formation.

Unraveling the interplay between hydrogen bonding and rotational energy barrier to fine-tune the properties of triazine molecular glasses

Mexylaminotriazine derivatives form molecular glasses with outstanding glass-forming ability (GFA), glass kinetic stability (GS), and tunable glass transition temperature. This work establishes key molecular parameters for efficient glass engineering.

New knowledge and tools for crystal design: local coordination versus overall network topology and much more

The problem of predicting crystal structures is discussed in the context of artificial intelligence systems.