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

Автоматическое проектирование химических соединений с использованием управляемого данными непрерывного представления молекул
Alán Aspuru‐Guzik, Rafael Gómez-Bombarelli, Jennifer N. Wei, David Duvenaud, José Miguel Hernández-Lobato, Benjamín Sánchez-Lengeling, Dennis Sheberla, Jorge Aguilera-Iparraguirre, Timothy D. Hirzel, Ryan P. Adams, Alán Aspuru-Guzik, Rafael Gómez‐Bombarelli, Jennifer N. Wei, David Duvenaud, José Miguel Hernández-Lobato, Dennis Sheberla, Jorge Aguilera‐Iparraguirre, Ryan Prescott Adams
2018-01-12

continuous molecular representationdeep neural networkdrug-like moleculeslatent space optimizationmolecular property prediction
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.
1
A deep neural network learns coupled encoder, decoder, and predictor functions that map molecular structures into and out of a multidimensional continuous latent space.
2
Latent-space representations support gradient-based optimization for efficiently searching chemical compounds with improved predicted properties.
3
The approach enables open-ended exploration and optimization across chemical compound spaces from discrete molecular representations.
4
The continuous representation enables generation of novel molecules through random-vector decoding, perturbation of known molecules, and interpolation between molecular structures.
5
The method is demonstrated on drug-like molecules and molecules containing fewer than nine heavy atoms.

drug-like molecules and molecules with fewer than nine heavy atoms

automatic generation, exploration, and gradient-based optimization of novel chemical structures using their continuous latent-space representations

Publication Details
Publication Date
2018-01-12
Journal
Publisher
ISSN
Cited by
4548
Access Type
Author Information
Authors
Alán Aspuru‐Guzik
Rafael Gómez-Bombarelli
Jennifer N. Wei
David Duvenaud
José Miguel Hernández-Lobato
Benjamín Sánchez-Lengeling
Dennis Sheberla
Jorge Aguilera-Iparraguirre
Timothy D. Hirzel
Ryan P. Adams
Alán Aspuru-Guzik
Rafael Gómez‐Bombarelli
Jennifer N. Wei
David Duvenaud
José Miguel Hernández-Lobato
Dennis Sheberla
Jorge Aguilera‐Iparraguirre
Ryan Prescott Adams
Explore further
Open the scid.ai AI chat with a ready-made request: it will find papers on a similar topic and help build a literature review.
Find similar papers in the chat →
Make a presentation
100%