XGraphBoost: Extracting Graph Neural Network-Based Features for a Better Prediction of Molecular Properties

Abstract

Determining the properties of chemical molecules is essential for screening candidates similar to a specific drug. These candidate molecules are further evaluated for their target binding affinities, side effects, target missing probabilities, etc. Conventional machine learning algorithms demonstrated satisfying prediction accuracies of molecular properties. A molecule cannot be directly loaded into a machine learning model, and a set of engineered features needs to be designed and calculated from a molecule. Such hand-crafted features rely heavily on the experiences of the investigating researchers. The concept of graph neural networks (GNNs) was recently introduced to describe the chemical molecules. The features may be automatically and objectively extracted from the molecules through various types of GNNs, e.g., GCN (graph convolution network), GGNN (gated graph neural network), DMPNN (directed message passing neural network), etc. However, the training of a stable GNN model requires a huge number of training samples and a large amount of computing power, compared with the conventional machine learning strategies. This study proposed the integrated framework XGraphBoost to extract the features using a GNN and build an accurate prediction model of molecular properties using the classifier XGBoost. The proposed framework XGraphBoost fully inherits the merits of the GNN-based automatic molecular feature extraction and XGBoost-based accurate prediction performance. Both classification and regression problems were evaluated using the framework XGraphBoost. The experimental results strongly suggest that XGraphBoost may facilitate the efficient and accurate predictions of various molecular properties. The source code is freely available to academic users at https://github.com/chenxiaowei-vincent/XGraphBoost.git.

Assisting Multitargeted Ligand Affinity Prediction of Receptor Tyrosine Kinases Associated Nonsmall Cell Lung Cancer Treatment with Multitasking Principal Neighborhood Aggregation

“…The training-validation loss ratio could serve as a heuristic to indicate overfitting in some instances, what constitutes a suitable threshold may differ according to the model type and the dataset. Various machine-learning models, especially in intricate architectures such as deep learning, have been found to be a practical approach, even when the ratio between training loss and validation loss is high [47][48][49]. A well-established phenomenon in deep learning, as well as some classical machine learning, has addressed this issue regarding the bias-variance tradeoff, known for the double descent risk curve [50].…”

Graph Neural Networks as a Potential Tool in Improving Virtual Screening Programs

“…Extensive literature reports concerning the benchmarks of algorithm models using the aforementioned databases applied to VS related tasks are available, such as molecular property predictions, fingerprint generation or the evaluation of structural protein-ligand docking parameters. These include the following: Support Vector Machine (SVM), Extreme Gradient Boost (XGBoost), Random Forest (RF), and Deep Neural Networks (DNN) ( Jiang et al, 2021 ) as representatives of descriptor-based models and many graph-based algorithm variants, such as MPNN—Message Passing Neural Networks ( Yang et al, 2019 ; Deng et al, 2021 ; Jiang et al, 2021 ) and networks implementing algorithm model variants involving spatial graph convolution, like GCN—Graph Convolution Network ( Li et al, 2017 ; Xiong et al, 2020 ; Menke and Koch, 2020 ; Deng et al, 2021 ; Hsieh et al, 2020 ) or GC—Graph Convolution ( Wu et al, 2018 ) and spectral graph convolution, such as AGCN–Adaptive Graph Convolution ( Li et al, 2018 ), graph based networks including attention mechanisms of interaction between near nodes or edges, i.e., AFP—Attentive Fingerprint ( Xiong et al, 2020 ; Jiang et al, 2021 ), PAGTN—Path-Augmented Graph Transformer Network ( Chen et al, 2019 ), EAGCN—Edge Attention GCN ( Shang et al, 2018 ), among others ( Wu et al, 2018 ; Lim et al, 2019 ).…”

“…Direct Message Passage Neural Network, D-MPNN—a graph based model combined with Extreme Gradient Boost, XGBoost—a descriptor-based network as the output layer, achieved the best results for several of the presented dataset ( Deng et al, 2021 ). Furthermore, the concatenation of molecular fingerprint vectors generated by conventional models with descriptors generated using graph models have been reported as providing the best prediction results when submitted to the final parameter generation layers ( Wang et al, 2019 ).…”

Improving Small Molecule pKa Prediction Using Transfer Learning with Graph Neural Networks

“…Another aspect of their attractiveness for molecular property prediction is the ease with which a molecule can be described as an undirected graph, transforming atoms to nodes and bonds to edges encoded both atom and bond properties. GNNs have proven to be useful and powerful tools in the machine learning molecular modeling toolbox [19,20].…”