Enhanced Graph Isomorphism Network for Molecular ADMET Properties Prediction

Abstract

The evaluation of absorption, distribution, metabolism, exclusion, and toxicity (ADMET) properties plays a key role in a variety of domains including industrial chemicals, agrochemicals, cosmetics, environmental science, food chemistry, and particularly drug development. Since molecules are often intrinsically described as molecular graphs, graph neural networks have recently been studied to improve the prediction of ADMET properties. Among many graph neural networks published in recent years, Graph Isomorphism Network (GIN) is a relatively recent and very promising one. In this paper, we propose an enhanced GIN, called MolGIN, via exploiting the bond features and differences influence of the atom neighbors to end-to-end predict ADMET properties. Based on GIN, MolGIN concatenates the bond feature together with node feature in the feature aggregator and applies a gate unit to adjust the atomic neighborhood weights to map the differences in the interaction strength between the central atom and its neighbors, such that more meaningful structural patterns of molecules can be explored toward better molecular modeling. Extensive experiments were conducted on seven public datasets to evaluate MolGIN against four baseline models with benchmark metrics. Experimental results of MolGIN were also compared with state-of-the-art results published in the last three years on each dataset. Experimental results in terms of RMSE and AUC show that MolGIN significantly boosts the prediction performance of GIN and markedly outperforms the baseline models, and achieves comparable or superior performance to state-of-the-art results.

Activated sludge models at the crossroad of artificial intelligence—A perspective on advancing process modeling

“…Several studies have systematically analyzed the identifiability of such models 1,27,28 including our own studies, considering typical plant data collected from intensive measurement campaigns, which showed that among 60 plus model parameters, only a handful of them (6-10 parameter subsets) could be uniquely estimated from the data. These issues has been recognized already by calibration protocols in fact 29,30 . The rest of the model parameters need to be fixed or assumed when applying these models to simulate the activated sludge plants.…”

“…More details of these different molecular featurizations and models can be found elsewhere 29 . The graph-based featurizations and neural networks have recently gained significant research attention in cheminformatics and bioinformatics due to their superior performances on molecular ML tasks, as found in recent literature 29,30,60 . For instance, a spatial graph-based molecular representation of amino acid residue pairs has allowed AlfaFold 61 , an AI program developed by DeepMind, to perform 3D protein structure predictions far more accurately than ever made.…”

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

“…For instance, the nonidentical molecules, decalin and 1,1-bicyclopentane (Figure S1 in the Supplementary Materials), can be recognized as the same molecule in molecular graph network due to their identical topology [34]. This issue resulted in a lower predictive power of the model [35], causing a difficulty with bioactivity prediction [36]. Accordingly, multiple aggregation strategies have been proposed to improve the GNN's performance, leading to the development of the principal neighborhood aggregation (PNA) method by DeepMind [37].…”

Improving Molecular Graph Neural Network Explainability with Orthonormalization and Induced Sparsity

“…Graph convolutional neural networks (GCNNs) have shown particular promise in predicting properties of small molecules. These properties can be biological activity against protein targets (Yang et al, 2020;Sakai et al, 2021;Nguyen et al, 2020), more general pharmacokinetics and physicochemical properties (Montanari et al, 2019;Peng et al, 2020;Feinberg et al, 2020) or toxicity (Ma et al, 2020). Using accurate in silico predictions of such properties to prioritize the synthesis of compounds can lead to tremendous time and cost savings during drug discovery projects.…”