Conditional Molecular Design with Deep Generative Models

Seokho Kang, Kyunghyun Cho

Research output: Contribution to journalArticle

Abstract

Although machine learning has been successfully used to propose novel molecules that satisfy desired properties, it is still challenging to explore a large chemical space efficiently. In this paper, we present a conditional molecular design method that facilitates generating new molecules with desired properties. The proposed model, which simultaneously performs both property prediction and molecule generation, is built as a semisupervised variational autoencoder trained on a set of existing molecules with only a partial annotation. We generate new molecules with desired properties by sampling from the generative distribution estimated by the model. We demonstrate the effectiveness of the proposed model by evaluating it on drug-like molecules. The model improves the performance of property prediction by exploiting unlabeled molecules and efficiently generates novel molecules fulfilling various target conditions.

Original languageEnglish (US)
Pages (from-to)43-52
Number of pages10
JournalJournal of Chemical Information and Modeling
Volume59
Issue number1
DOIs
StatePublished - Jan 28 2019

Fingerprint

Molecules
drug
learning
performance
Learning systems
Sampling
Pharmaceutical Preparations

ASJC Scopus subject areas

  • Chemistry(all)
  • Chemical Engineering(all)
  • Computer Science Applications
  • Library and Information Sciences

Cite this

Conditional Molecular Design with Deep Generative Models. / Kang, Seokho; Cho, Kyunghyun.

In: Journal of Chemical Information and Modeling, Vol. 59, No. 1, 28.01.2019, p. 43-52.

Research output: Contribution to journalArticle

@article{f6a7637d20b141d5a37b4577dfdd7ca7,
title = "Conditional Molecular Design with Deep Generative Models",
abstract = "Although machine learning has been successfully used to propose novel molecules that satisfy desired properties, it is still challenging to explore a large chemical space efficiently. In this paper, we present a conditional molecular design method that facilitates generating new molecules with desired properties. The proposed model, which simultaneously performs both property prediction and molecule generation, is built as a semisupervised variational autoencoder trained on a set of existing molecules with only a partial annotation. We generate new molecules with desired properties by sampling from the generative distribution estimated by the model. We demonstrate the effectiveness of the proposed model by evaluating it on drug-like molecules. The model improves the performance of property prediction by exploiting unlabeled molecules and efficiently generates novel molecules fulfilling various target conditions.",
author = "Seokho Kang and Kyunghyun Cho",
year = "2019",
month = "1",
day = "28",
doi = "10.1021/acs.jcim.8b00263",
language = "English (US)",
volume = "59",
pages = "43--52",
journal = "Journal of Chemical Information and Modeling",
issn = "1549-9596",
publisher = "American Chemical Society",
number = "1",

}

TY - JOUR

T1 - Conditional Molecular Design with Deep Generative Models

AU - Kang, Seokho

AU - Cho, Kyunghyun

PY - 2019/1/28

Y1 - 2019/1/28

N2 - Although machine learning has been successfully used to propose novel molecules that satisfy desired properties, it is still challenging to explore a large chemical space efficiently. In this paper, we present a conditional molecular design method that facilitates generating new molecules with desired properties. The proposed model, which simultaneously performs both property prediction and molecule generation, is built as a semisupervised variational autoencoder trained on a set of existing molecules with only a partial annotation. We generate new molecules with desired properties by sampling from the generative distribution estimated by the model. We demonstrate the effectiveness of the proposed model by evaluating it on drug-like molecules. The model improves the performance of property prediction by exploiting unlabeled molecules and efficiently generates novel molecules fulfilling various target conditions.

AB - Although machine learning has been successfully used to propose novel molecules that satisfy desired properties, it is still challenging to explore a large chemical space efficiently. In this paper, we present a conditional molecular design method that facilitates generating new molecules with desired properties. The proposed model, which simultaneously performs both property prediction and molecule generation, is built as a semisupervised variational autoencoder trained on a set of existing molecules with only a partial annotation. We generate new molecules with desired properties by sampling from the generative distribution estimated by the model. We demonstrate the effectiveness of the proposed model by evaluating it on drug-like molecules. The model improves the performance of property prediction by exploiting unlabeled molecules and efficiently generates novel molecules fulfilling various target conditions.

UR - http://www.scopus.com/inward/record.url?scp=85050348242&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85050348242&partnerID=8YFLogxK

U2 - 10.1021/acs.jcim.8b00263

DO - 10.1021/acs.jcim.8b00263

M3 - Article

VL - 59

SP - 43

EP - 52

JO - Journal of Chemical Information and Modeling

JF - Journal of Chemical Information and Modeling

SN - 1549-9596

IS - 1

ER -