IP Library › Granted Patent US 12,217,834
Granted Patent B2
US 12,217,834 · App. 17/614,856 · Granted Feb 4, 2025

Molecular graph generation from structural features using an artificial neural network

Inventors: Paul Maragakis (New York, NY); Hunter Nisonoff (New York, NY); Peter Skopp (New York, NY); John Salmon (New York, NY)
Assignee: D. E. Shaw Research, LLC
G16C20/30G06N3/045G06N3/08G16C20/50G16C20/70G16C20/80
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,217,834
App. No.
17/614,856
Filed
Nov 29, 2021
Granted
Feb 4, 2025
Kind
B2
Examiner
DO, AN H
Art Unit
2853
USPC
702/27
Abstract

Discovering molecules (which may be known or may never have been cataloged or ever synthesized) that have desired characteristics is addressed using a machine learning approach. As compared to a brute-force search of a database of known molecules, which may not be computationally feasible, the present machine learning approach renders identification of both known and unknown molecules computationally tractable. Furthermore, the computational effort is largely shifted to training of the machine learning system using a database of known molecules, and the generation of molecules to match any particular characteristics requires relatively little computation. The molecules using the present approach may be further studied, for example, with computer-based simulation or after physical synthesis using biological experimentation to ultimately yield useful chemical compounds.

Claims (59)

1. A computer-implemented method for generating a molecular graph comprising:

obtaining a data representation of first structural features of a first molecule;

processing, using a computer, the data representation of the first structural features of the first molecule to yield a first latent representation of the first molecule; and

processing, using the computer, the first latent representation to yield a data representation of a molecular graph of at least one molecule matching the first structural features;

wherein the first structural features comprises conformational properties of a conformation of a molecule.

2. The method of claim 1 , further comprising providing the molecular graph of the at least one molecule for evaluation of chemical or biological properties of said at least one molecule.

3. The method of claim 2 , further comprising evaluating a molecule represented by the molecular graph by at least one of:

(a) computer-implemented molecular dynamics simulation,

(b) computer-implemented docking evaluation,

(c) free energy calculation to characterize expected binding behavior,

(d) physical synthesis based on a known synthesis procedure or based on a computer-determined synthesis procedure, and

(e) evaluation of the molecule in chemical or biological experiments.

4. The method of claim 1 , wherein the first molecule is a known reference molecule.

5. The method of claim 1 , wherein obtaining the data representation of the first structural features includes processing a data representation of a first molecular graph to yield the first structural features.

6. The method of claim 5 , wherein processing the data representation of the first molecular graph comprises applying a plurality of rules to yield the first structural features.

7. The method of claim 1 , wherein the first molecule is an unknown target molecule.

8. The method of claim 1 , wherein the data representation of the molecular graph of the at least one molecule comprises a linear symbolic representation.

9. The method of claim 8 , wherein the linear symbolic representation corresponds to a SMILES representation.

10. The method of claim 8 , wherein the linear symbolic representation comprises symbols each representing at least some individual atoms and symbols each representing a group of bonded atoms.

11. The method of claim 8 , wherein the linear symbolic representation comprises a compression of the SMILES representation.

12. The method of claim 1 , wherein the conformational properties are determined for a known reference molecule.

13. The method of claim 1 , wherein the conformational properties are determined as desired in an unknown target molecule.

14. The method of claim 1 , wherein the representation of conformational properties comprises features tied to three-dimensional locations of those attributes in a conformation of a molecule.

15. The method of claim 14 , wherein the conformational properties comprise a set of locations and discrete categories of features of a molecule at said locations.

16. The method of claim 14 wherein the conformational properties comprise properties of a plurality of low-energy conformations of a molecule.

17. The method of claim 1 , wherein processing the data representation of the first structural features to yield the first latent representation comprises using a first artificial neural network.

18. The method of claim 1 , wherein processing the first latent representation to yield the data representation of the molecular graph of at least one molecule matching the first structural features comprises processing the latent representation using a second artificial neural network to yield a sequence representation of the at least one molecule matching the structural features.

19. The method of claim 1 , wherein the first structural features comprise a fingerprint determined from a molecular graph.

20. The method of claim 19 , wherein the fingerprint comprises a fixed length binary vector.

21. The method of claim 20 , wherein the fingerprint representation comprises an extended connectivity fingerprint.

22. The method of claim 19 wherein the fingerprint representation is encoded to a continuous vector.

23. The method of claim 1 , wherein the molecular graph of the at least one molecule has the first structural features.

24. A computer-implemented method for generating a molecular graph comprising:

obtaining a data representation of first structural features of a first molecule;

processing, using a computer, the data representation of the first structural features of the first molecule to yield a first latent representation of the first molecule; and

processing, using the computer, the first latent representation to yield a data representation of a molecular graph of at least one molecule matching the first structural features;

wherein processing the first latent representation to yield the data representation of the molecular graph of at least one molecule matching the first structural features comprises processing the latent representation using a second artificial neural network to yield a sequence representation of the at least one molecule matching the structural features; and

wherein processing the latent representation to yield a sequence representation of the at least one molecule matching the structural features comprises processing the latent representation to yield a plurality of sequence representations, determining structural features for each representation of said plurality of sequence representations, and comparing the determined structural features to determine a match of the determined structural features to the first structural features.

25. The method of claim 24 , wherein processing the latent representation to yield the plurality of sequence representations includes generating a first sequence representation, and wherein generating the first sequence representation includes generating successive distributions of next symbols in the sequence representation using the latent representation as an input to a second artificial neural network and searching possible sequences using the successive distributions to yield one or more best sequences.

26. The method of claim 25 , wherein generating the distribution of a next symbol includes using the latent representation and a prefix of the next symbol in the sequence as input to the second artificial neural network.

27. The method of claim 25 , wherein searching the possible sequences comprises performing an A-star search procedure.

28. A computer-implemented method of training a neural network for generating a molecular graph comprising:

obtaining a first set of molecular graphs;

processing each molecular graph of the first set of molecular graphs to determine respective sequence representations;

processing each molecular graph of the first set of molecular graphs to determine respective structural features; and

training, using the structural features and respective sequence representations, a combination of a first neural network and a second neural network;

wherein the first neural network implements a transformation from the structural features to a latent feature representation, and the second neural network implements a transformation from the latent feature representation to a sequence representation.

29. The method of claim 28 wherein the structural features comprise fingerprints.

30. The method of claim 28 , further comprising:

obtaining a second set of molecular graphs;

processing each molecular graph of the second set of molecular graphs to determine respective latent representations; and

processing each molecular graph of the second set of molecular graphs to determine respective conformational properties;

training, using the latent representations and the conformational properties, a third neural network;

wherein the third neural network implements a transformation from conformational properties to a corresponding latent representation; and

wherein a combination of the third neural network and the second neural network together implement a transformation from conformational properties to a sequence representation of a molecular graph.

31. The method of claim 28 , further comprising:

obtaining a set of pairs of molecular graphs, each pair having a first molecular graph and a second molecular graph, the second molecular graph representing a second molecule having a different degree of a property than a first molecule represented by the first molecular graph;

processing the first molecular graphs to determine corresponding first structural features; and

applying a transfer learning procedure to at least the second neural network using the pairs of first structural features and second molecular graphs.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Jul 29, 2022
From: MARAGAKIS, PAUL; NISONOFF, HUNTER; SKOPP, PETER; SALMON, JOHN K.
To: D.E. SHAW RESEARCH, LLC
Reel/Frame 060667/0790 →
Continuity (3)
Provisional Application 62855355 · May 31, 2019
Provisional Application 62855388 · May 31, 2019
Related Publication 20220230713A1 · Jul 21, 2022
References Cited (45)
US 11263534B1 · Prat · 2022 [cited by examiner]
US 11710049B2 · Tal · 2023 [cited by examiner]
US 20210287137A1 · Park · 2021 [cited by examiner]
Josep Arús-Pous, Thomas Blaschke, Silas Ulander, Jean-Louis Reymond, Hongming Chen, and Ola Engkvist. “Exploring the GDB-13 chemical space using deep generative models.” Journal of cheminformatics, 11(1):20, 2019. [cited by applicant]
Seth D Axen, Xi-Ping Huang, Elena L Caceres, Leo Gendelev, Bryan L Roth, and Michael J Keiser. “A simple representation of three-dimensional molecular structure.” Journal of medicinal chemistry, 60(17):7393-7409, 2017. [cited by applicant]
Hongming Chen, Ola Engkvist, Yinhai Wang, Marcus Olivecrona, and Thomas Blaschke. “The rise of deep learning in drug discovery.” Drug discovery today, 23(6):1241-1250, 2018. [cited by applicant]
Daniel C Elton, Zois Boukouvalas, Mark D Fuge, and Peter W Chung. “Deep learning for molecular design-a review of the state of the art.” arXiv preprint arXiv:1903.04388, 2019. [cited by applicant]
Jesse Engel, Matthew Hoffman, and Adam Roberts. “Latent constraints: Learning to Generate Conditionally from Unconditional Generative Models.” arXiv preprint arXiv:1711.05772, 2017. [cited by applicant]
Justin Gilmer, Samuel S Schoenholz, Patrick F Riley, Oriol Vinyals, and George E Dahl. “Neural Message Passing for Quantum Chemistry.” In Proceedings of the 34th International Conference on Machine Learning-Volume 70, p… [cited by applicant]
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, and Alán Aspuru-Guzik. “A… [cited by applicant]
Anvita Gupta, Alex T Müller, Berend JH Huisman, Jens A Fuchs, Petra Schneider, and Gisbert Schneider. “Generative Recurrent Networks for De Novo Drug Design.” Molecular informatics, 37(1-2):1700111, 2018. [cited by applicant]
Sepp Hochreiter and Jürgen Schmidhuber. “Long short-term memory.” Neural computation, 9(8):1735-1780, 1997. [cited by applicant]
Wengong Jin, Regina Barzilay, and Tommi Jaakkola. “Junction Tree Variational Autoencoder for Molecular Graph Generation.” arXiv preprint arXiv:1802.04364, 2018. [cited by applicant]
Anonymous Authors. “Learning Multimodal Graph-to-Graph Translation for Molecular Optimization.” arXiv preprint arXiv:1812.01070, 2018. [cited by applicant]
Artur Kadurin, Sergey Nikolenko, Kuzma Khrabrov, Alex Aliper, and Alex Zhavoronkov. “druGAN: An Advanced Generative Adversarial Autoencoder Model for de Novo Generation of New Molecules with Desired Molecular Properties… [cited by applicant]
Yujia Li, Oriol Vinyals, Chris Dyer, Razvan Pascanu, and Peter Battaglia. “Learning Deep Generative Models of Graphs.” arXiv preprint arXiv:1803.03324, 2018. [cited by applicant]
Yu-Chen Lo, Stefano E Rensi, Wen Torng, and Russ B Altman. “Machine learning in chemoinformatics and drug discovery.” Drug Discovery Today, vol. 00, No. 00, 2018, pp. 1-9. [cited by applicant]
Hans Matter and Thorsten Pötter. “Comparing 3D Pharmacophore Triplets and 2D Fingerprints for Selecting Diverse Compound Subsets.” Journal of Chemical Information and Computer Sciences, 39(6):1211-1225, 1999. [cited by applicant]
Stephen Merity, Nitish Shirish Keskar, and Richard Socher. “Regularizing and Optimizing LSTM Language Models.” arXiv preprint arXiv:1708.02182, 2017. [cited by applicant]
Marcus Olivercrona, Thomas Blaschke, Ola Engkvist, and Hongming Chen. “Molecular De-Novo Design through Deep Reinforcement Learning.” Journal of cheminformatics, 9(1):48, 2017. [cited by applicant]
Jane Panteleev, Hua Gao, and Lei Jia. “Recent applications of machine learning in medicinal chemistry.” Bioorganic & Medicinal Chemistry Letters, 2018. [cited by applicant]
Mariya Popova, Olexandr Isayev, and Alexander Tropsha. “Deep reinforcement learning for de novo drug design.” Science Advances, 4(7):eaap7885, 2018. [cited by applicant]
Raquel Rodriguez-Perez, Tomoyuki Miyao, Swarit Jasial, Martin Vogt, and Jürgen Bajorath. “Prediction of Compound Profiling Matrices Using Machine Learning.” ACS omega, 3(4):4713-4723, 2018. [cited by applicant]
David Rogers and Mathew Hahn. “Extended-Connectivity Fingerprints.” Journal of chemical information and modeling, 50(5):742-754, 2010. [cited by applicant]
Benjamin Sanchez-Lengeling and Alán Aspuru-Guzik. “Inverse molecular design using machine learning: Generative models for matter engineering.” Science, 361(6400): 360-365, 2018. [cited by applicant]
Marwin HS Segler, Thierry Kogej, Christian Tyrchan, and Mark P Waller. “Generating Focussed Molecule Libraries for Drug Discovery with Recurrent Neural Networks.” ACS central science, 4(1):120-131, 2017. [cited by applicant]
Marwin HS Segler, Mike Preuss, and Mark P Waller. “Planning chemical syntheses with deep neural networks and symbolic AI.” Nature, 555(7698):604, 2018. [cited by applicant]
Ilya Sutskever, Oriol Vinyals, and Quoc V. Le. “Sequence to Sequence Learning with Neural Networks.” In Advances in neural information processing systems, pp. 3104-3112, 2014. [cited by applicant]
Oriol Vinyals, Alexander Toshev, Samy Bengio, and Dumitru Erhan. “Show and Tell: A Neural Image Caption Generator.” In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 3156-3164, 2015. [cited by applicant]
Izhar Wallach, Michael Dzamba, and Abraham Heifets.“AtomNet: A Deep Convolutional Neural Network for Bioactivity Prediction in Structure-based Drug Discovery.” arXiv preprint arXiv:1510.02855, 2015. [cited by applicant]
W. Patrick Walters. “Virtual Chemical Libraries: Miniperspective.” Journal of Medicinal Chemistry, 62(3):1116-1124, 2018. [cited by applicant]
R Winter, F Montanari, A Steffen, H Briem, F Noé, and D Clevert. “Efficient Multi-Objective Molecular Optimization in a Continuous Latent Space.” ChemRxiv: 7971101, 2019. [cited by applicant]
Sam Wiseman and Alexander M Rush. “Sequence-to-Sequence Learning as Beam-Search Optimization.” arXiv preprint arXiv:1606.02960, 2016. [cited by applicant]
Dongyu Xue, Yukang Gong, Zhaoyi Yang, Guohui Chuai, Sheng Qu, Aizong Shen, Jing Yu, and Qi Liu. “Advances and challenges in deep generative models for de novo molecule generation.” Wiley Interdisciplinary Reviews: Compu… [cited by applicant]
Stanislaw Jastrzbski, Damian Lésniak, and Wojciech Marian Czarnecki. “Learning to SMILE(S).” arXiv preprint arXiv:1602.06289, 2016. [cited by applicant]
Peter Pogany, Navot Arad, Sam Genway, and Stephen D Pickett. “De Novo Molecule Design by Translating from Reduced Graphs to SMILES.” Journal of chemical information and modeling, 2018. [cited by applicant]
Evgeny Putin, Arip Asadulaev, Yan Ivanenkov, Vladimir Aladinskiy, Benjamin Sanchez-Lengeling, Alan Aspuru-Guzik, and Alex Zhavoronkov. “Reinforced Adversarial Neural Computer for de Novo Molecular Design.” Journal of Ch… [cited by applicant]
Winter, Frank Noé, Djork-Arné Clevert. “Neuraldecipher—Reverse-Engineering ECFP Fingerprints to Their Molecular Structures.” ChemRxiv, 2020. [cited by applicant]
Tuan Le, Robin Winter, Frank Noé, and Djork-Arné Clevert. “Neuraldecipher—Reverse-engineering ECFP Fingerprints to their molecular structures” Machine Learning Research, 2020. [cited by applicant]
Ilya Sutskever, Oriol Vinyals, Quoc V. Le. “Sequence to Sequence Learning with Neural Networks.” 2014. [cited by applicant]
Paul Maragakis et al. “A deep-learning view of chemical space designed to facilitate drug discovery.” Arxiv.org, Cornell University Library, 201 Olin Library Cornell University Ithaca, NY 14853, Feb. 7, 2020, XP08159414… [cited by applicant]
Kyungdoc Kim et al. “Deep-learning-based inverse design model for intelligent discovery of organic molecules”, NPJ Computational Materials, vol. 4, No. 1, Dec. 3, 2018, XP055633998, DOI: 10.1038/s41524-018-0128-1. [cited by applicant]
Daniil Polykovskiy et al. “Molecular Sets (MOSES): A Benchmarking Platform for Molecular Generation Models”, Arxiv.org, Cornell University Library, 201 Olin Library Cornell University Ithaca, NY 14853, Nov. 29, 2018, P0… [cited by applicant]
Yibo Li et al. “Multi-Objective De Novo Drug Design with Conditional Graph Generative Model,” Arxiv.org, Cornell University Library, 201 Olin Library Cornell University Ithaca, NY 14853, Jan. 18, 2018, XP081227289. [cited by applicant]
International Search Report and Written Opinion, PCT Application No. PCT/US2020/035137, mailed Oct. 7, 2020 (13 pages). [cited by applicant]