CoChat
Sign Up
Cited by 3,106Computational Drug Discovery MethodsMachine Learning in Materials ScienceHistory and advancements in chemistry

Automatic Chemical Design Using a Data-Driven Continuous Representation of Molecules

Open access
Gómez-Bombarelli, Rafael, Wei, Jennifer N., Duvenaud, David, Hernández-Lobato, José Miguel, Sánchez-Lengeling, Benjamín, Sheberla, Dennis, Aguilera-Iparraguirre, Jorge, Hirzel, Timothy D. +2 more·2018·ACS Central Science
DOI 10.1021/acscentsci.7b00572PMC PMC5833007
Open this paper in CoChat — free accountOpen PDF

Abstract

We report a method to convert discrete representations of molecules to and from a multidimensional continuous representation. This model allows us to generate new molecules for efficient exploration and optimization through open-ended spaces of chemical compounds. A deep neural network was trained on hundreds of thousands of existing chemical structures to construct three coupled functions: an encoder, a decoder, and a predictor. The encoder converts the discrete representation of a molecule into a real-valued continuous vector, and the decoder converts these continuous vectors back to discrete molecular representations. The predictor estimates chemical properties from the latent continuous vector representation of the molecule. Continuous representations of molecules allow us to automatically generate novel chemical structures by performing simple operations in the latent space, such as decoding random vectors, perturbing known chemical structures, or interpolating between molecules. Continuous representations also allow the use of powerful gradient-based optimization to efficiently guide the search for optimized functional compounds. We demonstrate our method in the domain of drug-like molecules and also in a set of molecules with fewer that nine heavy atoms.

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1021/acscentsci.7b00572

Cites

49 of 49 references verified0 citation errors · 0 not checkable

Statistics

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • To meet particular goals, optimization methods such as post-processing approaches (transfer learning and reinforcement learning) and Bayesian optimization are applied.

    “Furthermore, to meet particular goals, optimization methods are applied, including techniques such as post-processing approaches (e.g., transfer learning and reinforcement learning) and Bayesian optimization11.”

    From PCF-VAE: posterior collapse free variational autoencoder for de novo drug design (2025) · cites this paper as [11]

    The Introduction (page not available) describes optimizing molecules for desired properties and says gradient-based optimization can be combined with Bayesian inference. The Results and Discussion section, “Optimization of Molecules via Properties” (page not available), reports optimization using a Gaussian-process model. Our copy does not describe transfer learning or reinforcement learning as post-processing approaches, so the full list attributed to this paper is not supported. (LLM, conf=0.87)

    Experimental — accuracy not yet measured

  • These techniques can be applied to the latent space of models just as readily as the feature space, for example, to highlight regions of promise for enrichment or to reveal governing design principles.

    “These techniques can be applied to the latent space of models just as readily as the feature space, for example, to highlight regions of promise for enrichment (37,65) or to reveal governing design principles (64).”

    From Machine Learning for the Discovery, Design, and Engineering of Materials. (2022) · cites this paper as [64]

  • ML techniques including supervised learning, active learning, and generative models have been used to help synthesize, characterize, model, and design materials.

    “ML techniques including supervised learning, active learning, and generative models have been used to help synthesize, characterize, model, and design materials, as described in reviews [105,490] and more recent work [299].”

    From Tackling climate change with machine learning (2023) · cites this paper as [299]

  • ML has had recent success in predicting specific chemical, material, and optical properties without the need for expensive experimentation or brute-force simulation, including in aerosols.

    “ML has had recent success in predicting specific chemical, material, and optical properties without the need for expensive experimentation or brute-force simulation [105,299,490,653], including in aerosols [465,500].”

    From Tackling climate change with machine learning (2023) · cites this paper as [299]

  • Deep learning successes in computational biology include accurate contact prediction in proteins, improved diagnostics for several diseases, molecular design and simulation.

    “Deep learning has begun to permeate every scientific discipline, and in computational biology its successes include accurate contact prediction in proteins [61], improved diagnostics for several diseases [128], molecular design [129] and simulation [130,131].”

    From The prospects of quantum computing in computational molecular biology (2021) · cites this paper as [129]

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Cites

Flagged first, then by citations

Research works printed but not linked: 7

  1. Duvenaud D. K.; Maclaurin D.; Iparraguirre J.; Gómez-Bombarell R.; Hirzel T.; Aspuru-Guzik A.; Adams R. P. Convolutional Networks on Graphs for Learning Molecular Fingerprints. Adv. Neural Information Processing Syst. 2015, 2215– 2223.
  2. Aggarwal C. C.; Hinneburg A.; Keim D. A. Database Theory – ICDT 2001:8th International Conference London, UK, January 4–6, 2001 Proceedings; Springer: Berlin, Heidelberg, 2001; pp 420– 434.
  3. Rasmussen C. E.; Williams C. K. Gaussian Processes for Machine Learning; MIT Press: Cambridge, 2006; Vol. 1.
  4. Janz D.; van der Westhuizen J.; Hernández-Lobato J. M. Actively Learning what makes a Discrete Sequence Valid, 2017; http://arxiv.org/abs/1603.00856. · arXiv
  5. Jaques N.; Gu S.; Bahdanau D.; Hernández-Lobato J. M.; Turner R. E.; Eck D. Sequence tutor: Conservative fine-tuning of sequence generation models with kl-control. International Conference on Machine Learning, 2017; pp 1645– 1654.
  6. Sutskever I.; Vinyals O.; Le Q. V. Sequence to sequence learning with neural networks. Adv. Neural Information Processing Syst. 2014, 3104– 3112.
  7. Abadi M. TensorFlow: A system for large-scale machine learning. 12th USENIX Symposium on Operating Systems Design and Implementation (OSDI 16), 2016; pp 265– 283.

Cited by

Sorted by citations

Authors
Gómez-Bombarelli, R., Wei, J. N., Duvenaud, D., Hernández-Lobato, J. M., Sánchez-Lengeling, B., Sheberla, D., Aguilera-Iparraguirre, J., Hirzel, T. D., Adams, R. P., & Aspuru-Guzik, A.
Affiliations
  • Kyulux North America Inc., 10 Post Office Square, Suite 800, Boston, Massachusetts 02109, United States
  • Harvard University (US)
  • University of Toronto (CA)
  • University of Cambridge (GB)
  • Google (United States) (US)
  • Department of Chemistry and Chemical Biology, Harvard University, Cambridge, Massachusetts 02138, United States
ORCID
Venue
ACS Central Science
Publisher
American Chemical Society (ACS)
Published
2018-01-12
Volume
4
Issue
2
Pages
268-276
ISSN
  • 2374-7943 (ISSN-L)
  • 2374-7951
DOI
10.1021/acscentsci.7b00572
PMC
PMC5833007
OpenAlex
W2529996553
Cited by
3,106
References
49
Citations by year
2026: 430 · 2025: 463 · 2024: 463 · 2023: 431 · 2022: 366 · 2021: 385 · 2020: 289 · 2019: 185 · 2018: 81 · 2017: 11 · 2013: 1
FWCI
175.58
Type
journal article
Language
en
Open access
Licence
Locations
  • ACS Central Science (publishedVersion, other-oa)
  • Europe PMC (PubMed Central) (publishedVersion)
  • PubMed Central (publishedVersion)
  • PubMed (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • Digital Access to Scholarship at Harvard (DASH) (Harvard University) (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • OSTI OAI (U.S. Department of Energy Office of Scientific and Technical Information) (submittedVersion)
  • arXiv (Cornell University)
Funders
  • Basic Energy Sciences (DE-SC0015959)
  • Division of Graduate Education (DGE-1144152)
  • Alfred P. Sloan Foundation
  • Division of Information and Intelligent Systems (IIS-1421780)
  • Samsung Advanced Institute of Technology
Topics
  • Physical Sciences › Computer Science › Computational Theory and Mathematics › Computational Drug Discovery Methods
  • Physical Sciences › Materials Science › Materials Chemistry › Machine Learning in Materials Science
  • Physical Sciences › Chemistry › Physical and Theoretical Chemistry › History and advancements in chemistry
Keywords
molecular encoding, latent space, generative models, molecular property prediction, drug-like molecules, small molecules, gradient-based optimization

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review