2
|
Bort W, Baskin II, Gimadiev T, Mukanov A, Nugmanov R, Sidorov P, Marcou G, Horvath D, Klimchuk O, Madzhidov T, Varnek A. Discovery of novel chemical reactions by deep generative recurrent neural network. Sci Rep 2021; 11:3178. [PMID: 33542271 PMCID: PMC7862614 DOI: 10.1038/s41598-021-81889-y] [Citation(s) in RCA: 27] [Impact Index Per Article: 9.0] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 08/08/2020] [Accepted: 01/06/2021] [Indexed: 12/18/2022] Open
Abstract
The "creativity" of Artificial Intelligence (AI) in terms of generating de novo molecular structures opened a novel paradigm in compound design, weaknesses (stability & feasibility issues of such structures) notwithstanding. Here we show that "creative" AI may be as successfully taught to enumerate novel chemical reactions that are stoichiometrically coherent. Furthermore, when coupled to reaction space cartography, de novo reaction design may be focused on the desired reaction class. A sequence-to-sequence autoencoder with bidirectional Long Short-Term Memory layers was trained on on-purpose developed "SMILES/CGR" strings, encoding reactions of the USPTO database. The autoencoder latent space was visualized on a generative topographic map. Novel latent space points were sampled around a map area populated by Suzuki reactions and decoded to corresponding reactions. These can be critically analyzed by the expert, cleaned of irrelevant functional groups and eventually experimentally attempted, herewith enlarging the synthetic purpose of popular synthetic pathways.
Collapse
Affiliation(s)
- William Bort
- Laboratory of Chemoinformatics, UMR 7140 CNRS, University of Strasbourg, 1, rue Blaise Pascal, 67000, Strasbourg, France
| | - Igor I Baskin
- Laboratory of Chemoinformatics, UMR 7140 CNRS, University of Strasbourg, 1, rue Blaise Pascal, 67000, Strasbourg, France
- Laboratory of Chemoinformatics and Molecular Modeling, Butlerov Institute of Chemistry, Kazan Federal University, Kremlyovskaya str. 18, 420008, Kazan, Russia
- Department of Materials Science and Engineering, Technion - Israel Institute of Technology, 3200003, Haifa, Israel
| | - Timur Gimadiev
- Institute for Chemical Reaction Design and Discovery (WPI-ICReDD), Hokkaido University, Kita 21 Nishi 10, Kita-ku, Sapporo, 001-0021, Japan
| | - Artem Mukanov
- Laboratory of Chemoinformatics and Molecular Modeling, Butlerov Institute of Chemistry, Kazan Federal University, Kremlyovskaya str. 18, 420008, Kazan, Russia
| | - Ramil Nugmanov
- Laboratory of Chemoinformatics and Molecular Modeling, Butlerov Institute of Chemistry, Kazan Federal University, Kremlyovskaya str. 18, 420008, Kazan, Russia
| | - Pavel Sidorov
- Institute for Chemical Reaction Design and Discovery (WPI-ICReDD), Hokkaido University, Kita 21 Nishi 10, Kita-ku, Sapporo, 001-0021, Japan
| | - Gilles Marcou
- Laboratory of Chemoinformatics, UMR 7140 CNRS, University of Strasbourg, 1, rue Blaise Pascal, 67000, Strasbourg, France
| | - Dragos Horvath
- Laboratory of Chemoinformatics, UMR 7140 CNRS, University of Strasbourg, 1, rue Blaise Pascal, 67000, Strasbourg, France
| | - Olga Klimchuk
- Laboratory of Chemoinformatics, UMR 7140 CNRS, University of Strasbourg, 1, rue Blaise Pascal, 67000, Strasbourg, France
| | - Timur Madzhidov
- Laboratory of Chemoinformatics and Molecular Modeling, Butlerov Institute of Chemistry, Kazan Federal University, Kremlyovskaya str. 18, 420008, Kazan, Russia
| | - Alexandre Varnek
- Laboratory of Chemoinformatics, UMR 7140 CNRS, University of Strasbourg, 1, rue Blaise Pascal, 67000, Strasbourg, France.
- Institute for Chemical Reaction Design and Discovery (WPI-ICReDD), Hokkaido University, Kita 21 Nishi 10, Kita-ku, Sapporo, 001-0021, Japan.
| |
Collapse
|
3
|
Zankov DV, Madzhidov TI, Rakhimbekova A, Gimadiev TR, Nugmanov RI, Kazymova MA, Baskin II, Varnek A. Conjugated Quantitative Structure–Property Relationship Models: Application to Simultaneous Prediction of Tautomeric Equilibrium Constants and Acidity of Molecules. J Chem Inf Model 2019; 59:4569-4576. [DOI: 10.1021/acs.jcim.9b00722] [Citation(s) in RCA: 4] [Impact Index Per Article: 0.8] [Reference Citation Analysis] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/28/2022]
Affiliation(s)
- Dmitry V. Zankov
- Laboratory of Chemoinformatics and Molecular Modeling, Butlerov Institute of Chemistry, Kazan Federal University, Kremlyovskaya str. 18, 420008 Kazan, Russia
| | - Timur I. Madzhidov
- Laboratory of Chemoinformatics and Molecular Modeling, Butlerov Institute of Chemistry, Kazan Federal University, Kremlyovskaya str. 18, 420008 Kazan, Russia
| | - Assima Rakhimbekova
- Laboratory of Chemoinformatics and Molecular Modeling, Butlerov Institute of Chemistry, Kazan Federal University, Kremlyovskaya str. 18, 420008 Kazan, Russia
| | - Timur R. Gimadiev
- Laboratory of Chemoinformatics and Molecular Modeling, Butlerov Institute of Chemistry, Kazan Federal University, Kremlyovskaya str. 18, 420008 Kazan, Russia
| | - Ramil I. Nugmanov
- Laboratory of Chemoinformatics and Molecular Modeling, Butlerov Institute of Chemistry, Kazan Federal University, Kremlyovskaya str. 18, 420008 Kazan, Russia
| | - Marina A. Kazymova
- Laboratory of Chemoinformatics and Molecular Modeling, Butlerov Institute of Chemistry, Kazan Federal University, Kremlyovskaya str. 18, 420008 Kazan, Russia
| | - Igor I. Baskin
- Laboratory of Chemoinformatics and Molecular Modeling, Butlerov Institute of Chemistry, Kazan Federal University, Kremlyovskaya str. 18, 420008 Kazan, Russia
- Faculty of Physics, Moscow State University, Vorob’evy gory 1, 119234 Moscow, Russia
| | - Alexandre Varnek
- Laboratory of Chemoinformatics, UMR 7140 CNRS, University of Strasbourg, 1, rue Blaise Pascal, 67000 Strasbourg, France
- Institute for Chemical Reaction Design and Discovery (WPI-ICReDD), Hokkaido University, Kita 21 Nishi 10, Kita-ku, 001-0021 Sapporo, Japan
| |
Collapse
|
4
|
Abstract
Solutions of organic molecules containing one or more heterocycles with conjugated bonds may exist as a mixture of tautomers, but typically only a few of them are significantly populated even though the potential number grows combinatorially with the number of protonation and deprotonation sites. Generating the most stable tautomers from a given input structure is an important and challenging task, and numerous algorithms to tackle it have been proposed in the literature. This work describes a novel approach for tautomer prediction that involves the combined use of molecular mechanics, semiempirical quantum chemistry, and density functional theory. The key idea in our method is to identify the protonation and deprotonation sites using estimated micro-p Ka's for every atom in the molecule as well as in its nearest protonated and deprotonated forms. To generate tautomers in a systematic way with minimal bias, we then consider the full set of tautomers that arise from the combinatorial distribution of all such mobile protons among all protonatable sites, with efficient postprocessing to screen away high-energy species. To estimate the micro-p Ka's, we present a new method designed for the current task, but we emphasize that any alternative method can be used in conjunction with our basic algorithm. Our approach is therefore grounded in the computational prediction of physical properties in aqueous solution, in contrast to other approaches that may rely on the use of hard-coded rules of proton distribution, previously observed tautomerization patterns from a known chemical space, or human input. We present examples of the application of our algorithm to organic and drug-like molecules, with a focus on novel structures where traditional methods are expected to perform worse.
Collapse
Affiliation(s)
- Mark A Watson
- Schrödinger, Inc. , 120 West 45th Street , New York , New York 10036 , United States
| | - Haoyu S Yu
- Schrödinger, Inc. , 120 West 45th Street , New York , New York 10036 , United States
| | - Art D Bochevarov
- Schrödinger, Inc. , 120 West 45th Street , New York , New York 10036 , United States
| |
Collapse
|
5
|
Sattarov B, Baskin II, Horvath D, Marcou G, Bjerrum EJ, Varnek A. De Novo Molecular Design by Combining Deep Autoencoder Recurrent Neural Networks with Generative Topographic Mapping. J Chem Inf Model 2019; 59:1182-1196. [PMID: 30785751 DOI: 10.1021/acs.jcim.8b00751] [Citation(s) in RCA: 77] [Impact Index Per Article: 15.4] [Reference Citation Analysis] [Abstract] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Indexed: 11/28/2022]
Abstract
Here we show that Generative Topographic Mapping (GTM) can be used to explore the latent space of the SMILES-based autoencoders and generate focused molecular libraries of interest. We have built a sequence-to-sequence neural network with Bidirectional Long Short-Term Memory layers and trained it on the SMILES strings from ChEMBL23. Very high reconstruction rates of the test set molecules were achieved (>98%), which are comparable to the ones reported in related publications. Using GTM, we have visualized the autoencoder latent space on the two-dimensional topographic map. Targeted map zones can be used for generating novel molecular structures by sampling associated latent space points and decoding them to SMILES. The sampling method based on a genetic algorithm was introduced to optimize compound properties "on the fly". The generated focused molecular libraries were shown to contain original and a priori feasible compounds which, pending actual synthesis and testing, showed encouraging behavior in independent structure-based affinity estimation procedures (pharmacophore matching, docking).
Collapse
Affiliation(s)
- Boris Sattarov
- Laboratory of Chemoinformatics , UMR 7177 University of Strasbourg/CNRS , 4 rue B. Pascal , 67000 Strasbourg , France
| | - Igor I Baskin
- Faculty of Physics , M.V. Lomonosov Moscow State University , Leninskie Gory , Moscow 19991 , Russia
| | - Dragos Horvath
- Laboratory of Chemoinformatics , UMR 7177 University of Strasbourg/CNRS , 4 rue B. Pascal , 67000 Strasbourg , France
| | - Gilles Marcou
- Laboratory of Chemoinformatics , UMR 7177 University of Strasbourg/CNRS , 4 rue B. Pascal , 67000 Strasbourg , France
| | - Esben Jannik Bjerrum
- Wildcard Pharmaceutical Consulting, Zeaborg Science Center, Frødings Allé 41 , 2860 Søborg , Denmark
| | - Alexandre Varnek
- Laboratory of Chemoinformatics , UMR 7177 University of Strasbourg/CNRS , 4 rue B. Pascal , 67000 Strasbourg , France
| |
Collapse
|