Costless Performance Improvement in Machine Learning for Graph-Based Molecular Analysis

Abstract

Graph neural networks (GNNs) have attracted significant attention from the chemical science community because molecules can be represented as a featured graph. In particular, graph convolutional network (GCN) and its variants have been widely used and have shown a state-of-the-art performance in analyzing molecules, such as molecular label classification, drug discovery, and molecular property prediction. However, in molecular analysis, existing GCNs have two fundamental limitations: (1) information of the molecular scale is distorted and (2) global structures in a molecule are ignored. These limitations can seriously degrade the performance in the machine learning-based molecular analysis because the scale and global structure information of a molecule occasionally have a significant effect on the molecular properties. To overcome the limitations of existing GCNs, we comprehensively analyzed the structure of GCNs and developed a costless solution for the limitations of GCNs. To demonstrate the effectiveness of our solution, extensive experiments were conducted on various benchmark datasets.

Temporal and Spatial Differences of Urban Ecological Environment and Economic Development Based on Graph Neural Network

“…Aiming toward the limitations of existing graph neural networks, Na G S et al comprehensively analyze the structure of graph neural networks and develop a cost-free molecular analysis solution for the limitations of graph neural networks. Finally, they verified the usefulness as a solution by conducting experimentation with a variety of reference sets of data [ 1 ]. Lee L H. et al proposed a graph neural network model for medical applications.…”

Tuplewise Material Representation Based Machine Learning for Accurate Band Gap Prediction

“…43 The CGCNN gives a reliable prediction of materials properties with an accuracy comparable to DFT calculations, but its complex representation learning process and a large number of model parameters easily lead to overfitting and make the training difficult. 39,40 In addition, CGCNN only exploits atomic and bonding features, 38,44,45 not crystal-level properties (e.g., band gap, formation energy, bulk modulus, etc. ), because of their inherent network architectures, so that the feature vectors contain only a portion of the whole materials properties.…”

A review on machine learning approaches and trends in drug discovery

“…V is the set of atoms in the molecule, A corresponds to the adjacent matrix that indicates the connectivity between atoms, and the X matrix represents the atomic characteristics for each atom. Therefore, each graph is mathematically represented as G = (V, A, X) [49] , [50] .…”