검색 상세

Toward 3D Structure Augmented Deep Molecular Generation

  • 박진준 (아주대학교 일반대학원)

초록/요약

Which molecular generation method is best for large molecular generations? Finding a good lead molecule is an important task in drug discovery. Recently several deep graph generative models have been developed for generating novel molecules that can be further tested for synthesizability in the drug development process. Most of the developed models are trained on small molecules with a maximum length of thirty. However, there is a need for the generation of larger molecules. We tested six recently proposed graph neural network-based molecular generation methods on their large molecular generation performance using two datasets from the LigandBox database, which contain larger molecules than typically used ZINC250k and QM9 datasets. In addition, we propose a modified model using 3D coordinate information of molecules and evaluate this model together with recent models. We use twelve evaluation measures to evaluate the quality of the generated molecules, including stability measures such as logP values and QEDs.

more

목차

1 Introduction 1
2 Related Works 4
2.1 Graph Neural Networks 4
2.2 Deep Graph Models for Molecular Generation 4
2.2.1 JT-VAE 5
2.2.2. HierVAE 6
2.2.3 GraphAF 6
2.2.4 GraphDF 7
2.3.5 GraphEBM 7
2.3.6 MoFlow 8
3 Method 9
3.1 Expended GraphEBM 9
3.2 Dataset 11
3.3 Data Sampling Process 12
3.3.1 Evaluation Metrics 13
3.4 Compared Models 14
4 Results 15
4.1 Random Generation Performance Evaluation 15
4.2 Property Optimization Performance Evaluation 17
5 Conclusion 23
Reference 24

more
반출 Meta View 목록