Masked Graph Modeling for Molecule Generation

Mahmood, Omar; Mansimov, Elman; Bonneau, Richard; Cho, Kyunghyun

doi:10.26434/chemrxiv.13143167.v3

Cited by 3 publications

(3 citation statements)

References 38 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…These language models can be used to generate molecular libraries for drug discovery 6 or built into variational autoencoders (VAE) 3 , 7 where bayesian optimization can be used to search through the model’s latent space for drug-like molecules. Other models generate molecules as graphs either sequentially 8 – 14 using graph neural networks 15 , 16 or generate whole molecules in one shot 17 – 20 . Two of the most popular: CGAVE and JTVAE can be directly constrained to enforce valency restrictions.…”

Section: Introductionmentioning

confidence: 99%

Language models can learn complex molecular distributions

2022

View full text Add to dashboard Cite

Deep generative models of molecules have grown immensely in popularity, trained on relevant datasets, these models are used to search through chemical space. The downstream utility of generative models for the inverse design of novel functional compounds, depends on their ability to learn a training distribution of molecules. The most simple example is a language model that takes the form of a recurrent neural network and generates molecules using a string representation. Since their initial use, subsequent work has shown that language models are very capable, in particular, recent research has demonstrated their utility in the low data regime. In this work, we investigate the capacity of simple language models to learn more complex distributions of molecules. For this purpose, we introduce several challenging generative modeling tasks by compiling larger, more complex distributions of molecules and we evaluate the ability of language models on each task. The results demonstrate that language models are powerful generative models, capable of adeptly learning complex molecular distributions. Language models can accurately generate: distributions of the highest scoring penalized LogP molecules in ZINC15, multi-modal molecular distributions as well as the largest molecules in PubChem. The results highlight the limitations of some of the most popular and recent graph generative models– many of which cannot scale to these molecular distributions.

show abstract

Section: Introductionmentioning

confidence: 99%

Language models can learn complex molecular distributions

2022

View full text Add to dashboard Cite

show abstract

“…Code, pretrained MGM models, training and generation scripts for MGM and baseline models, and lists of generated molecules can be found at https://github.com/nyu-dl/ dl4chem-mgm 66 .…”

Section: Data Availabilitymentioning

confidence: 99%

Masked graph modeling for molecule generation

et al. 2021

View full text Add to dashboard Cite

De novo, in-silico design of molecules is a challenging problem with applications in drug discovery and material design. We introduce a masked graph model, which learns a distribution over graphs by capturing conditional distributions over unobserved nodes (atoms) and edges (bonds) given observed ones. We train and then sample from our model by iteratively masking and replacing different parts of initialized graphs. We evaluate our approach on the QM9 and ChEMBL datasets using the GuacaMol distribution-learning benchmark. We find that validity, KL-divergence and Fréchet ChemNet Distance scores are anti-correlated with novelty, and that we can trade off between these metrics more effectively than existing models. On distributional metrics, our model outperforms previously proposed graph-based approaches and is competitive with SMILES-based approaches. Finally, we show our model generates molecules with desired values of specified properties while maintaining physiochemical similarity to the training distribution.

show abstract

“…Deep learning (DL) generative model offers a promising approach for accessing the chemical space. Molecular generative models can learn the distribution of a set of known chemical compounds and output new molecules in a highly efficient and reliable manner [10][11][12][13][14][15] . These models are essentially ligand-based, as they rely on the prior knowledge of known actives to design similar molecules that potentially bind to the same target with the possibility to fine-tune their physicochemical properties 16,17 .…”

Section: Introductionmentioning

confidence: 99%

A Simple Way to Incorporate Target Structural Information in Molecular Generative Models

Zhang

Huang

2023

Preprint

View full text Add to dashboard Cite

Deep learning generative models are now being applied in various fields including drug discovery. In this work, we propose a novel approach to include target 3D structural information in molecular generative models for structure-based drug design. The method combines a message-passing neural network model that predicts docking scores with a generative neural network model as its reward function to navigate the chemical space searching for molecules that bind favorably with a specific target. A key component of the method is to construct target-specific molecular sets for training to overcome potential transferability issues. Tests on eight target proteins showed a 100-fold increase in hit generation compared to conventional docking calculations, and the ability to generate molecules similar to approved drugs or known active ligands for specific targets without prior knowledge. This method provides a general and highly efficient solution for structure-based molecular generation.

show abstract

Masked Graph Modeling for Molecule Generation

Cited by 3 publications

References 38 publications

Language models can learn complex molecular distributions

Language models can learn complex molecular distributions

Masked graph modeling for molecule generation

A Simple Way to Incorporate Target Structural Information in Molecular Generative Models

Contact Info

Product

Resources

About