Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2020 Dec 18:11:565644.
doi: 10.3389/fphar.2020.565644. eCollection 2020.

Molecular Sets (MOSES): A Benchmarking Platform for Molecular Generation Models

Daniil Polykovskiy  1 , Alexander Zhebrak  1 , Benjamin Sanchez-Lengeling  2 , Sergey Golovanov  3 , Oktai Tatanov  3 , Stanislav Belyaev  3 , Rauf Kurbanov  3 , Aleksey Artamonov  3 , Vladimir Aladinskiy  1 , Mark Veselov  1 , Artur Kadurin  1 , Simon Johansson  4 , Hongming Chen  4 , Sergey Nikolenko  1   3   5 , Alán Aspuru-Guzik  6   7   8   9 , Alex Zhavoronkov  1
Affiliations

Molecular Sets (MOSES): A Benchmarking Platform for Molecular Generation Models

Daniil Polykovskiy et al. Front Pharmacol. .

Abstract

Generative models are becoming a tool of choice for exploring the molecular space. These models learn on a large training dataset and produce novel molecular structures with similar properties. Generated structures can be utilized for virtual screening or training semi-supervized predictive models in the downstream tasks. While there are plenty of generative models, it is unclear how to compare and rank them. In this work, we introduce a benchmarking platform called Molecular Sets (MOSES) to standardize training and comparison of molecular generative models. MOSES provides training and testing datasets, and a set of metrics to evaluate the quality and diversity of generated structures. We have implemented and compared several molecular generation models and suggest to use our results as reference points for further advancements in generative chemistry research. The platform and source code are available at https://github.com/molecularsets/moses.

Keywords: benchmark; deep learning; distribution learning; drug discovery; generative models.

PubMed Disclaimer

Conflict of interest statement

DP, AZhe, VA, MV, and AZha work for Insilico Medicine, a commercial artificial intelligence company. SG, OT, SB, RK, AA, and SN work for Neuromation OU, a company engaged in AI development through synthetic data and generative models. SJ and HC work for a pharmaceutical company AstraZeneca. AA-G is a cofounder and board member of, and consultant for, Kebotix, an artificial intelligence-driven molecular discovery company and a member of the science advisory board of Insilico Medicine.

Figures

FIGURE 1
FIGURE 1
Molecular Sets (MOSES) pipeline. The open-source library provides a dataset, baseline models, and evaluation metrics.
FIGURE 2
FIGURE 2
Different views on a vanillin molecule.
FIGURE 3
FIGURE 3
Examples of molecules from MOSES dataset.
FIGURE 4
FIGURE 4
Distribution of chemical properties for MOSES dataset and sets of generated molecules. In brackets—Wasserstein-1 distance to MOSES test set. Parameters: molecular weight, octanol-water partition coefficient (logP), quantitative estimation of drug-likeness (QED) and synthetic accessibility score (SA).
See this image and copyright information in PMC

References

    1. Aliper A., Plis S., Artemov A., Ulloa A., Mamoshina P., Zhavoronkov A. (2016). Deep learning applications for predicting pharmacological properties of drugs and drug repurposing using transcriptomic data. Mol. Pharm. 13, 2524–2530. 10.1021/acs.molpharmaceut.6b00248 - DOI - PMC - PubMed
    1. Arús-Pous J., Johansson S. V., Prykhodko O., Bjerrum E. J., Tyrchan C., Reymond J.-L., et al. (2019). Randomized smiles strings improve the quality of molecular generative models. J. Cheminf. 11, 1–13. 10.1186/s13321-019-0393-0 - DOI - PMC - PubMed
    1. Baell J. B., Holloway G. A. (2010). New substructure filters for removal of pan assay interference compounds (PAINS) from screening libraries and for their exclusion in bioassays. J. Med. Chem. 53, 2719–2740. 10.1021/jm901137j - DOI - PubMed
    1. Bemis G. W., Murcko M. A. (1996). The properties of known drugs. 1. molecular frameworks. J. Med. Chem. 39, 2887–2893. 10.1021/jm9602928 - DOI - PubMed
    1. Benhenda M. (2017). ChemGAN challenge for drug discovery: can AI reproduce natural chemical diversity? Available from: https://arxiv.org/abs/1708.08227.

LinkOut - more resources