Evaluation of Deep Learning Architectures for Aqueous Solubility Prediction
Оценка архитектур глубокого обучения для прогнозирования растворимости в воде
2022-04-25
SCID: 54.1/ykh2vxd8
Discuss with AI
SchNetaqueous solubility predictiondeep learninggraph neural networksmolecular descriptors
Figures from the paper
Abstract (AI)
Determining the aqueous solubility of molecules is a vital step in many pharmaceutical, environmental, and energy storage applications. Despite efforts made over decades, there are still challenges associated with developing a solubility prediction model with satisfactory accuracy for many of these applications. The goals of this study are to assess current deep learning methods for solubility prediction, develop a general model capable of predicting the solubility of a broad range of organic molecules, and to understand the impact of data properties, molecular representation, and modeling architecture on predictive performance. Using the largest currently available solubility data set, we implement deep learning-based models to predict solubility from the molecular structure and explore several different molecular representations including molecular descriptors, simplified molecular-input line-entry system strings, molecular graphs, and three-dimensional atomic coordinates using four different neural network architectures-fully connected neural networks, recurrent neural networks, graph neural networks (GNNs), and SchNet. We find that models using molecular descriptors achieve the best performance, with GNN models also achieving good performance. We perform extensive error analysis to understand the molecular properties that influence model performance, perform feature analysis to understand which information about the molecular structure is most valuable for prediction, and perform a transfer learning and data size study to understand the impact of data availability on model performance.
Key Findings
1
Extensive error and feature analyses identify molecular properties and structural information that influence solubility-prediction performance.
2
Molecular descriptors provide the best predictive performance among the tested representations, while graph neural networks also achieve good performance.
3
The comparison covers descriptors, SMILES strings, molecular graphs, and three-dimensional atomic coordinates with fully connected, recurrent, graph, and SchNet architectures.
4
The study evaluates deep learning approaches for predicting aqueous solubility across a broad range of organic molecules using the largest currently available solubility dataset.
5
Transfer-learning and data-size analyses assess how data availability affects model performance and generalization.
Research Object
Aqueous solubility of organic molecules
Research Subject
Deep-learning prediction accuracy and the effects of molecular representations, neural-network architectures, molecular properties, and data availability
Publication Details
Publication Date
2022-04-25
Journal
Publisher
ISSN
Cited by
96
Open access PDF
Access Type
Author Information
Download PDF
Subscribe to digest