Search

Highlights from the article: After confirming the reasonable performance of our GCN model, we deleted each one from the 12 atom features and trained the feature-deleted models with the same protocol as before in order to understand the contribution of individual atom features to the aqueous solubility.

Although deep‐learning (DL) models suggest unprecedented prediction capabilities in tackling various chemical problems, their demonstrated tasks have so far been limited to the scalar properties including the magnitude of vectorial properties, such as molecular dipole moments. A rotation‐equivariant MolNet_Equi model, proposed in this paper, understands and recognizes the molecular rotation in the 3D Euclidean space, and exhibits the ability to predict directional dipole moments in the rotation‐sensitive mode, as well as showing superior performance for the prediction of scalar properties. Three consecutive operations of molecular rotation RM ${\left(R\left(M\right)\right)}$ , dipole‐moment prediction φμRM ${\left({\phi{} }_{\mu }\left(R\left(M\right)\right)\right)}$ , and dipole‐moment inverse‐rotation R-1φμRM ${\left({R}^{-1}\left({\phi{} }_{\mu }\left(R\left(M\right)\right)\right)\right)}$ do not alter the original prediction of the total dipole moment of a molecule φμM ${\left({\phi{} }_{\mu }\right(M\left)\right)}$ , assuring the rotational equivariance of MolNet_Equi. Furthermore, MolNet_Equi faithfully predicts the absolute direction of dipole moments given molecular poses, albeit the model has been trained only with the information on dipole‐moment magnitudes, not directions. This work highlights the potential of incorporating fundamental yet crucial chemical rules and concepts into DL models, leading to the development of chemically intuitive models. [ABSTRACT FROM AUTHOR]

Most graph neural networks (GNNs) in deep‐learning chemistry collect and update atom and molecule features from the fed atom (and, in some cases, bond) features, basically based on the two‐dimensional (2D) graph representation of 3D molecules. However, the 2D‐based models do not faithfully represent 3D molecules and their physicochemical properties, exemplified by the overlooked field effect that is a "through‐space" effect, not a "through‐bond" effect. We propose a GNN model, denoted as MolNet, which accommodates the 3D non‐bond information in a molecule, via a noncovalent adjacency matrix A‾ , and also bond‐strength information from a weighted bond matrix B. Comparative studies show that MolNet outperforms various baseline GNN models and gives a state‐of‐the‐art performance in the classification task of BACE dataset and regression task of ESOL dataset. This work suggests a future direction for the construction of deep‐learning models that are chemically intuitive and compatible with the existing chemistry concepts and tools. [ABSTRACT FROM AUTHOR]

This work proposes the data augmentation by molecular rotation, with consideration that the protein‐ligand binding events are rotation‐variant. As a proof‐of‐concept, known active (i. e., 1‐labeled) ligands to human β‐secretase 1 (BACE‐1) are rotated for the generation of 0‐labeled data, and the rotation‐dependent prediction accuracy of 3D graph convolutional network (3DGCN) is investigated after data augmentation. The data augmentation makes the orientation‐recognizing ability of 3DGCN improved significantly in the classification task for BACE‐1/ligand binding. Furthermore, the data‐augmented 3DGCN has a capability for predicting active ligands from a candidate dataset, via improved performance of orientation recognition, which would be applied to virtual drug screening and discovery. [ABSTRACT FROM AUTHOR]

This work proposes the use of IN−A (IN: identity matrix; A: adjacency matrix), instead of IN+A, the normalized form of which has intensively been used for the construction of graph convolutional networks (GCNs), in deep‐learning chemistry. The performance of the GCN model with D−1/2IN−AD−1/2 in its convolution step is at least on a par with the vanilla GCN that uses D~−1/2IN+AD~−1/2 (D~: degree matrix of IN+A) in various chemistry datasets, such as FreeSolv, ESOL, lipophilicity, and blood–brain barrier penetration datasets. It could be seen that the use of IN−A might be more chemically intuitive than the use of IN+A, potentially embracing the information on bond properties, such as dipole moment, and functional groups in a molecule. This work suggests unavoidable necessity of tackling molecular‐representation problems in deep‐learning chemistry from unprecedented angles of view for advanced development and construction of chemically intuitive deep‐learning models. [ABSTRACT FROM AUTHOR]

3D graph convolutional network (3DGCN) is a graph neural network (GNN) variant that utilizes the 3D bond information on molecules for chemical tasks. In 3DGCN, the pooling operation extracts the molecular features from the atomic features for molecule embedding and dimensionality reduction. In this work, we investigated the pooling effects on the performance of 3DGCN in the classification of protein–ligand binding events, especially when a ligand is rotated with respect to a target protein in the 3D Cartesian coordinate. 3DGCN showed the general ability for recognizing the ligand rotation, and its prediction accuracy was found to be pooling‐dependent. For example, 3DGCN with max pooling did not recognize the rotations of known active ligands for human β‐secretase 1 faithfully, compared with the other pooling operations (sum, avg, and set2set). This work would contribute to augmented architecture evolution of 3DGCN for the chemical tasks that require the 3D molecular information including chirality. [ABSTRACT FROM AUTHOR]

Development of deep-learning models for intermolecular noncovalent (NC) interactions between proteins and ligands has great potential in the chemical and pharmaceutical tasks, including structure–activity relationship and drug design. It still remains an open question how to convert the three-dimensional, structural information of a protein–ligand complex into a graph representation in the graph neural networks (GNNs). It is also difficult to know whether a trained GNN model learns the NC interactions properly. Herein, we propose a GNN architecture that learns two distinct graphs—one for the intramolecular covalent bonds in a protein and a ligand, and the other for the intermolecular NC interactions between the protein and the ligand—separately by the corresponding covalent and NC convolutional layers. The graph separation has some advantages, such as independent evaluation on the contribution of each convolutional step to the prediction of dissociation constants, and facile analysis of graph-building strategies for the NC interactions. In addition to its prediction performance that is comparable to that of a state-of-the art model, the analysis with an explainability strategy of layer-wise relevance propagation shows that our model successfully predicts the important characteristics of the NC interactions, especially in the aspect of hydrogen bonding, in the chemical interpretation of protein–ligand binding. [ABSTRACT FROM AUTHOR]

Front Cover: Rotational Variance-Based Data Augmentation in 3D Graph Convolutional Network (Chem. Keywords: Data augmentation; Deep learning; 3D Graph convolutional network; Protein-ligand binding; Rotational variance EN Data augmentation Deep learning 3D Graph convolutional network Protein-ligand binding Rotational variance 2568 2568 1 09/22/21 20210920 NES 210920 B Data augmentation for 3D graph convolutional network (3DGCN) b has been achieved, for the classification task of human -secretase 1 (BACE-1)-ligand binding, by randomly rotating known active (1-labeled) ligands in the 3D Cartesian coordinate and assigning the rotated ones as label-0. Data augmentation, Deep learning, 3D Graph convolutional network, Protein-ligand binding, Rotational variance. [Extracted from the article]