Could graph neural networks learn better molecular representation for drug discovery? A comparison study of descriptor-based and graph-based models

Abstract

Graph neural networks (GNN) has been considered as an attractive modelling method for molecular property prediction, and numerous studies have shown that GNN could yield more promising results than traditional descriptor-based methods. In this study, based on 11 public datasets covering various property endpoints, the predictive capacity and computational efficiency of the prediction models developed by eight machine learning (ML) algorithms, including four descriptor-based models (SVM, XGBoost, RF and DNN) and four graph-based models (GCN, GAT, MPNN and Attentive FP), were extensively tested and compared. The results demonstrate that on average the descriptor-based models outperform the graph-based models in terms of prediction accuracy and computational efficiency. SVM generally achieves the best predictions for the regression tasks. Both RF and XGBoost can achieve reliable predictions for the classification tasks, and some of the graph-based models, such as Attentive FP and GCN, can yield outstanding performance for a fraction of larger or multi-task datasets. In terms of computational cost, XGBoost and RF are the two most efficient algorithms and only need a few seconds to train a model even for a large dataset. The model interpretations by the SHAP method can effectively explore the established domain knowledge for the descriptor-based models. Finally, we explored use of these models for virtual screening (VS) towards HIV and demonstrated that different ML algorithms offer diverse VS profiles. All in all, we believe that the off-the-shelf descriptor-based models still can be directly employed to accurately predict various chemical endpoints with excellent computability and interpretability.

Keywords: ADME/T prediction; Deep learning; Ensemble learning; Extreme gradient boosting; Graph neural networks.

Fig. 1

The general workflow of GNN in molecular property prediction

Fig. 2

Importance of the representative molecular descriptors (the top 20) and the corresponding SHAP values given by XGBoost for the a ESOL and b BBBP datasets. One molecule gets one dot on each descriptor’s line and dots stack up to show density

Fig. 3

The distributions of the prediction scores for the 1960 screened molecules predicted by the four descriptor-based models including a SVM, b XGBoost, c RF, d DNN and the four graph-based models including e GCN, f GAT, g MPNN and h Attentive FP

Fig. 4

The heat map of the Euclidean distances of the prediction scores for different model pairs

Fig. 5

The structural features of the potential inhibitors given by the four descriptor-based models including a SVM, b XGBoost, c RF and d DNN

Fig. 6

The structural features of the potential inhibitors predicted by the four graph-based models including a GCN, b GAT, c (MPNN) and d Attentive FP; e the structure of the known HIV inhibitor identified by all the eight models