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

Abstract

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 .

Explainable Graph Neural Networks with Data Augmentation for Predicting pKa of C–H Acids

“…Message-passing neural networks (MPNNs), as a variant of GNNs, − are capable of taking both nodes and edges features as inputs. By incorporating edges features into the model, it is possible to enhance the understanding and representation of molecular graphs, resulting in significantly improved accuracy in predicting molecular chemical properties. − Several MPNN-based network architectures have been proposed for predicting other compound attributes, such as cocrystal density, infrared spectra, bond order, and bond energy …”

“…By incorporating edges features into the model, it is possible to enhance the understanding and representation of molecular graphs, resulting in significantly improved accuracy in predicting molecular chemical properties. 32−35 Several MPNN-based network architectures have been proposed for predicting other compound attributes, such as cocrystal density, 36 infrared spectra, 37 bond order, and bond energy. 38 In this work, a model for predicting the pK a values of C−H acid based on the MPNN framework was constructed.…”

Attention Mechanism-Based Graph Neural Network Model for Effective Activity Prediction of SARS-CoV-2 Main Protease Inhibitors: Application to Drug Repurposing as Potential COVID-19 Therapy

“…28 Guo et al developed a general MPNNbased framework coupling with global attention to predict cocrystal density and further identified significant atoms to realize the interpretability of the model. 29 To accelerate the drug repurposing and discovery research, Wang et al presented a deep fusion anatomical therapeutic chemical (ATC) prediction model DeepATC, where GCN was used to extract drug topological information. 30 Pham et al developed a mechanism-driven neural network-based architecture DeepCE by incorporating GNN and multihead attention mechanism to support virtual screening of phenotype compounds.…”

“…Ryu et al employed GAT to accurately predict molecular polarity, solubility, energy, and additionally detected essential features directly relating to target properties . Guo et al developed a general MPNN-based framework coupling with global attention to predict cocrystal density and further identified significant atoms to realize the interpretability of the model . To accelerate the drug repurposing and discovery research, Wang et al presented a deep fusion anatomical therapeutic chemical (ATC) prediction model DeepATC, where GCN was used to extract drug topological information .…”

Speeding Up the Cocrystallization Process: Machine Learning-Combined Methods for the Prediction of Multicomponent Systems

“…For these methods, it was estimated a variable accuracy in the range of 30–80% depending on the API . To overcome the poor accuracy of the property-based methods, a combination of different tools was also proposed, showing an improvement in the coformer selection of specific systems. ,− Recently, data-driven ML approaches have become increasingly popular due to the rapidity of calculation and promising predictive accuracy. − Several algorithms were evaluated, such as support vector machine (SVM), random forest (RF), neural networks (NN), and partial least squares-discriminant analysis (PLS-DA), and also, a wide variety of molecular representations were considered, including molecular descriptors, fingerprint vectors, and molecular graphs . To mention a few studies, Fornari et al proposed using QSAR descriptors and the PLS-DA model to discriminate between the formation of cocrystals and physical mixtures .…”