Ligandformer: A Graph Neural Network for Predicting Compound Property with Robust Interpretation

Abstract

Robust and efficient interpretation of QSAR methods is quite useful to validate AI prediction rationales with subjective opinion (chemist or biologist expertise), understand sophisticated chemical or biological process mechanisms, and provide heuristic ideas for structure optimization in pharmaceutical industry. For this purpose, we construct a multi-layer self-attention based Graph Neural Network framework, namely Ligandformer, for predicting compound property with interpretation. Ligandformer integrates attention maps on compound structure from different network blocks. The integrated attention map reflects the machine's local interest on compound structure, and indicates the relationship between predicted compound property and its structure. This work mainly contributes to three aspects: 1. Ligandformer directly opens the black-box of deep learning methods, providing local prediction rationales on chemical structures. 2. Ligandformer gives robust prediction in different experimental rounds, overcoming the ubiquitous prediction instability of deep learning methods. 3. Ligandformer can be generalized to predict different chemical or biological properties with high performance. Furthermore, Ligandformer can simultaneously output specific property score and visible attention map on structure, which can support researchers to investigate chemical or biological property and optimize structure efficiently. Our framework outperforms over counterparts in terms of accuracy, robustness and generalization, and can be applied in complex system study.

Target-aware Molecule Generation for Drug Design Using a Chemical Language Model*

“…These compounds were then screened using molecular docking and Lig-andformer, an AI model for phenotypic activity prediction [52] (see Methods for details). At this stage, we eliminated the compounds with worse docking scores compared to Bortezomib and inactive compounds predicted by Ligand-former.…”

“…We utilized an adapted version of the Graph Neural Network (GNN) model as proposed in [52] to predict potential phenotypic activity. Compared with traditional GNNs, our model is designed such that the output from one layer is propagated to all subsequent layers for enhanced processing.…”

Artificial Intelligence in Drug Toxicity Prediction: Recent Advances, Challenges, and Future Perspectives

“…Recently, Guo et al used the GNN framework to build a multilayer self-attention method called Ligandformer for compound property prediction, such as aqueous solubility, Caco-2 cell permeability, and Ames mutagenesis. 119 On a data set of 7,617 compounds, their Ames prediction model made highly accurate predictions with a ROC-AUC of 92%. Hung and Gini developed a mutagenicity prediction model using a GCNN based on 2D molecular graphs without descriptors.…”

“…The most recent in silico mutagenicity prediction studies have utilized the strengths of large data sets and achieved high ACC ranging from 79% to 85% (Table ). Many ML and DL algorithms, such as RF, SVM, k-NN, XGB, GBT, GNN, and GCNN, have been used to build predictive models of mutagenicity instead of in vitro tests. ,,,,,− Chu et al used eight ML algorithms, including RF, SVM, XGB, partial least-squares discriminant analysis (PLSDA), mixture discriminant analysis (MDA), SVM, k-NN, and C5, to predict the mutagenicity of 7,617 compounds from a previous study . By combining a variety of molecular fingerprints and physicochemical molecular properties as compound descriptors with a selection of alternative modeling algorithms, models with good predictive ability were found that offered molecular insights and revealed aspects of molecules that cause mutagenicity.…”

“…The DNN-based prediction model outperformed traditional methods such as RF, SVM, and k-NN, with a prediction ACC of 83.1% and a ROC-AUC of 83.8% for the test set. Recently, Guo et al used the GNN framework to build a multilayer self-attention method called Ligandformer for compound property prediction, such as aqueous solubility, Caco-2 cell permeability, and Ames mutagenesis . On a data set of 7,617 compounds, their Ames prediction model made highly accurate predictions with a ROC-AUC of 92%.…”