Generating Focused Molecule Libraries for Drug Discovery with Recurrent Neural Networks
In de novo drug design, computational strategies are used to generate novel molecules with good affinity to the desired biological target. In this work, we show that recurrent neural networks can be trained as generative models for molecular structures, similar to statistical language models in natu...
| Main Authors: | , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
American Chemical Society
2017-12-01
|
| Series: | ACS Central Science |
| Online Access: | http://dx.doi.org/10.1021/acscentsci.7b00512 |
| _version_ | 1818506889916317696 |
|---|---|
| author | Marwin H. S. Segler Thierry Kogej Christian Tyrchan Mark P. Waller |
| author_facet | Marwin H. S. Segler Thierry Kogej Christian Tyrchan Mark P. Waller |
| author_sort | Marwin H. S. Segler |
| collection | DOAJ |
| description | In de novo drug design, computational strategies are used to generate novel molecules with good affinity to the desired biological target. In this work, we show that recurrent neural networks can be trained as generative models for molecular structures, similar to statistical language models in natural language processing. We demonstrate that the properties of the generated molecules correlate very well with the properties of the molecules used to train the model. In order to enrich libraries with molecules active toward a given biological target, we propose to fine-tune the model with small sets of molecules, which are known to be active against that target. Against Staphylococcus aureus, the model reproduced 14% of 6051 hold-out test molecules that medicinal chemists designed, whereas against Plasmodium falciparum (Malaria), it reproduced 28% of 1240 test molecules. When coupled with a scoring function, our model can perform the complete de novo drug design cycle to generate large sets of novel molecules for drug discovery. |
| first_indexed | 2024-12-10T22:10:48Z |
| format | Article |
| id | doaj.art-7a39575145a141cf9f675a73d24123e3 |
| institution | Directory Open Access Journal |
| issn | 2374-7943 2374-7951 |
| language | English |
| last_indexed | 2024-12-10T22:10:48Z |
| publishDate | 2017-12-01 |
| publisher | American Chemical Society |
| record_format | Article |
| series | ACS Central Science |
| spelling | doaj.art-7a39575145a141cf9f675a73d24123e32022-12-22T01:31:36ZengAmerican Chemical SocietyACS Central Science2374-79432374-79512017-12-014112013110.1021/acscentsci.7b00512Generating Focused Molecule Libraries for Drug Discovery with Recurrent Neural NetworksMarwin H. S. Segler0Thierry Kogej1Christian Tyrchan2Mark P. Waller3Institute of Organic Chemistry & Center for Multiscale Theory and Computation, Westfälische Wilhelms-Universität Münster, Münster, GermanyHit Discovery, Discovery Sciences, AstraZeneca R&D, Gothenburg, SwedenDepartment of Medicinal Chemistry, IMED RIA, AstraZeneca R&D, Gothenburg, SwedenDepartment of Physics & International Centre for Quantum and Molecular Structures, Shanghai University, Shanghai, ChinaIn de novo drug design, computational strategies are used to generate novel molecules with good affinity to the desired biological target. In this work, we show that recurrent neural networks can be trained as generative models for molecular structures, similar to statistical language models in natural language processing. We demonstrate that the properties of the generated molecules correlate very well with the properties of the molecules used to train the model. In order to enrich libraries with molecules active toward a given biological target, we propose to fine-tune the model with small sets of molecules, which are known to be active against that target. Against Staphylococcus aureus, the model reproduced 14% of 6051 hold-out test molecules that medicinal chemists designed, whereas against Plasmodium falciparum (Malaria), it reproduced 28% of 1240 test molecules. When coupled with a scoring function, our model can perform the complete de novo drug design cycle to generate large sets of novel molecules for drug discovery.http://dx.doi.org/10.1021/acscentsci.7b00512 |
| spellingShingle | Marwin H. S. Segler Thierry Kogej Christian Tyrchan Mark P. Waller Generating Focused Molecule Libraries for Drug Discovery with Recurrent Neural Networks ACS Central Science |
| title | Generating Focused Molecule Libraries for Drug Discovery with Recurrent Neural Networks |
| title_full | Generating Focused Molecule Libraries for Drug Discovery with Recurrent Neural Networks |
| title_fullStr | Generating Focused Molecule Libraries for Drug Discovery with Recurrent Neural Networks |
| title_full_unstemmed | Generating Focused Molecule Libraries for Drug Discovery with Recurrent Neural Networks |
| title_short | Generating Focused Molecule Libraries for Drug Discovery with Recurrent Neural Networks |
| title_sort | generating focused molecule libraries for drug discovery with recurrent neural networks |
| url | http://dx.doi.org/10.1021/acscentsci.7b00512 |
| work_keys_str_mv | AT marwinhssegler generatingfocusedmoleculelibrariesfordrugdiscoverywithrecurrentneuralnetworks AT thierrykogej generatingfocusedmoleculelibrariesfordrugdiscoverywithrecurrentneuralnetworks AT christiantyrchan generatingfocusedmoleculelibrariesfordrugdiscoverywithrecurrentneuralnetworks AT markpwaller generatingfocusedmoleculelibrariesfordrugdiscoverywithrecurrentneuralnetworks |