Dual graph convolutional neural network for predicting chemical networks.

BACKGROUND: Predicting of chemical compounds is one of the fundamental tasks in bioinformatics and chemoinformatics, because it contributes to various applications in metabolic engineering and drug discovery. The recent rapid growth of the amount of available data has enabled applications of computational approaches such as statistical modeling and machine learning method. Both a set of chemical interactions and chemical compound structures are represented as graphs, and various graph-based approaches including graph convolutional neural networks have been successfully applied to chemical network prediction. However, there was no efficient method that can consider the two different types of graphs in an end-to-end manner. RESULTS: We give a new formulation of the chemical network prediction problem as a link prediction problem in a graph of graphs (GoG) which can represent the hierarchical structure consisting of compound graphs and an inter-compound graph. We propose a new graph convolutional neural network architecture called dual graph convolutional network that learns compound representations from both the compound graphs and the inter-compound network in an end-to-end manner. CONCLUSIONS: Experiments using four chemical networks with different sparsity levels and degree distributions shows that our dual graph convolution approach achieves high prediction performance in relatively dense networks, while the performance becomes inferior on extremely-sparse networks.

Compound-protein interaction prediction with end-to-end learning of neural networks for graphs and sequences.

Motivation: In bioinformatics, machine learning-based methods that predict the compound-protein interactions (CPIs) play an important role in the virtual screening for drug discovery. Recently, end-to-end representation learning for discrete symbolic data (e.g. words in natural language processing) using deep neural networks has demonstrated excellent performance on various difficult problems. For the CPI problem, data are provided as discrete symbolic data, i.e. compounds are represented as graphs where the vertices are atoms, the edges are chemical bonds, and proteins are sequences in which the characters are amino acids. In this study, we investigate the use of end-to-end representation learning for compounds and proteins, integrate the representations, and develop a new CPI prediction approach by combining a graph neural network (GNN) for compounds and a convolutional neural network (CNN) for proteins. Results: Our experiments using three CPI datasets demonstrated that the proposed end-to-end approach achieves competitive or higher performance as compared to various existing CPI prediction methods. In addition, the proposed approach significantly outperformed existing methods on an unbalanced dataset. This suggests that data-driven representations of compounds and proteins obtained by end-to-end GNNs and CNNs are more robust than traditional chemical and biological features obtained from databases. Although analyzing deep learning models is difficult due to their black-box nature, we address this issue using a neural attention mechanism, which allows us to consider which subsequences in a protein are more important for a drug compound when predicting its interaction. The neural attention mechanism also provides effective visualization, which makes it easier to analyze a model even when modeling is performed using real-valued representations instead of discrete features. Availability and implementation: https://github.com/masashitsubaki. Supplementary information: Supplementary data are available at Bioinformatics online.

Quantum Deep Field: Data-Driven Wave Function, Electron Density Generation, and Atomization Energy Prediction and Extrapolation with Machine Learning.

Deep neural networks (DNNs) have been used to successfully predict molecular properties calculated based on the Kohn-Sham density functional theory (KS-DFT). Although this prediction is fast and accurate, we believe that a DNN model for KS-DFT must not only predict the properties but also provide the electron density of a molecule. This Letter presents the quantum deep field (QDF), which provides the electron density with an unsupervised but end-to-end physics-informed modeling by learning the atomization energy on a large-scale dataset. QDF performed well at atomization energy prediction, generated valid electron density, and demonstrated extrapolation.

Quantum Deep Descriptor: Physically Informed Transfer Learning from Small Molecules to Polymers.

In this study, we propose a physically informed transfer learning approach for materials informatics (MI) using a quantum deep descriptor (QDD) obtained from the quantum deep field (QDF). The QDF is a machine learning model based on density functional theory (DFT) and can be trained with a large database of molecular properties. The pre-trained QDF model can provide an effective molecular descriptor that encodes the fundamental quantum-chemical characteristics (i.e., the wave function or orbital, electron density, and energies of a molecule) learned from the large database; we refer to this descriptor as a QDD. We show that a QDD pre-trained with certain properties of small molecules can predict different properties (e.g., the band gap and dielectric constant) of polymers compared with some existing descriptors. We believe that our DFT-based, physically informed transfer learning approach will not only be useful for practical applications in MI but will also provide quantum-chemical insights into materials in the future. All codes used in this study are available at https://github.com/masashitsubaki.

Analysis and Usage: Subject-to-subject Linear Domain Adaptation in sEMG Classification.

Before the operation of a biosignal-based application, long-duration calibration is required to adjust the pre-trained classifier to a new user data (target data). For reducing such time-consuming step, linear domain adaptation (DA) transfer learning approaches, which transfer pooled data (source data) related to the target data, are highlighted. In the last decade, they have been applied to surface electromyogram (sEMG) data with the implicit assumption that sEMG data are linear. However, sEMGs typically have non-linear characteristics, and due to the discrepancy between the assumption and actual characteristics, linear DA approaches would cause a negative transfer. This study investigated how the correlation between the source and target data affects an 8-class forearm movement classification after applying linear DA approaches. As a result, we found significant positive correlations between the classification accuracy and the source-target correlation. Additionally, the source-target correlation depended on the motion class. Therefore, our results suggest that we should choose a non-linear DA approach when the source-target correlation among subjects or motion classes is low.

Fast and Accurate Molecular Property Prediction: Learning Atomic Interactions and Potentials with Neural Networks.

The discovery of molecules with specific properties is crucial to developing effective materials and useful drugs. Recently, to accelerate such discoveries with machine learning, deep neural networks (DNNs) have been applied to quantum chemistry calculations based on the density functional theory (DFT). While various DNNs for quantum chemistry have been proposed, these networks require various chemical descriptors as inputs and a large number of learning parameters to model atomic interactions. In this paper, we propose a new DNN-based molecular property prediction that (i) does not depend on descriptors, (ii) is more compact, and (iii) involves additional neural networks to model the interactions between all the atoms in a molecular structure. In the consideration of the molecular structure, we also model the potentials between all the atoms; this allows the neural networks to simultaneously learn the atomic interactions and potentials. We emphasize that these atomic "pair" interactions and potentials are characterized using the global molecular structure, a function of the depth of the neural networks; this leads to the implicit or indirect consideration of atomic "many-body" interactions and potentials within the DNNs. In the evaluation of our model with the benchmark QM9 data set, we achieved fast and accurate prediction performances for various quantum chemical properties. In addition, we analyzed the effects of learning the interactions and potentials on each property. Furthermore, we demonstrated an extrapolation evaluation, i.e., we trained a model with small molecules and tested it with large molecules. We believe that insights into the extrapolation evaluation will be useful for developing more practical applications in DNN-based molecular property predictions.