Learning Graph-Level Representation for Drug Discovery

By Junying Li, Deng Cai, Xiaofei He. 2017

125 citations recorded by the source corpus.

Predicating macroscopic influences of drugs on human body, like efficacy and toxicity, is a central problem of small-molecule based drug discovery. Molecules can be represented as an undirected graph, and we can utilize graph convolution networks to predication molecular properties. However, graph convolutional networks and other graph neural networks all focus on learning node-level representation rather than graph-level representation. Previous works simply sum all feature vectors for all nodes in the graph to obtain the graph feature vector for drug predication. In this paper, we introduce a dummy super node that is connected with all nodes in the graph by a directed edge as the representation of the graph and modify the graph operation to help the dummy super node learn graph-level feature. Thus, we can handle graph-level classification and regression in the same way as node-level classification and regression. In addition, we apply focal loss to address class imbalance in drug datasets. The experiments on MoleculeNet show that our method can effectively improve the performance of molecular properties predication.

Subfields: Artificial Intelligence, Computational Theory and Mathematics, Materials Chemistry, Graph Learning, Computational Chemistry & Drug Discovery

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computational Theory and Mathematics

In Materials Chemistry

In Graph Learning

In Computational Chemistry & Drug Discovery

Citations in the corpus

Strategies for Pre-training Graph Neural Networks (2019, methodology): “On the HIV dataset, where a number of recent works (Wu et al., 2018; Li et al., 2017; Ishiguro et al., 2019) have reported performance on the same scaffold split and using the same protocol, our best pre-trained model (ContextPred + Supervised) achieves state-of-the-art performance.”
Open Graph Benchmark: Datasets for Machine Learning on Graphs (2020, background): “, additional nodes that are connected to all nodes in the original graphs (Gilmer et al., 2017; Li et al., 2017; Ishiguro et al., 2019).”
All SMILES Variational Autoencoder (2019, result): “(a) ZINC250k M ODEL MAE LOG P MAE QED ECFP [53] 0.38 0.045 CVAE [16] 0.15 0.054 CVAE ENC [16] 0.13 0.037 G RAPH C ONV [12] [38] 0.854 P OTENTIAL N ET [14] 0.857 ± 0.006 T OXIC B LEND [71] 0.862 All SMILES 0.871”
Do Transformers Really Perform Badly for Graph Representation? (2021, methodology): “Besides the superior expressiveness than popular GNNs, we also find an interesting connection between using self-attention and the virtual node heuristic [15, 31, 24, 22].”
Geometric deep learning on molecular representations (2021, background): “Since their introduction as a means to predict quantum chemical properties of small molecules at the density functional theory (DFT) level [5], GNNs have found many applications in quantum chemistry [63–67], drug discovery [29, 68, 69], CASP [31], and molecular property prediction [70, 71].”
Do Transformers Really Perform Bad for Graph Representation? (2021, methodology): “Besides the superior expressiveness than popular GNNs, we also find an interesting connection between using self-attention and the virtual node heuristic [15, 30, 23, 21].”
Could graph neural networks learn better molecular representation for drug discovery? A comparison study of descriptor-based and graph-based models (2020, methodology): “In most of these reported studies, traditional ML models such as LR, RF, SVM (especially ‘gold standard’ RF) [31, 37] were employed to develop the prediction models based on a set of individual fingerprints (especially Extended Connectivity Fingerprints, ECFP) [31–33].”
The rise of deep learning in drug discovery. (2018, background): “[50], Xu et al. [51], Li et al. [52] and Coley et al .”
Graph convolutional networks for computational drug development and discovery (2019, methodology): “[32] introduced an alternative approach to generate graph-level representation using GCN.”
Graph Warp Module: an Auxiliary Module for Boosting the Power of Graph Neural Networks in Molecular Graph Analysis (2019, methodology): “In this study, we use supernode [8, 15, 20] to promote the global propagation of the information in molecular graphs.”