De Novo Design of Anti-COVID Drugs Using Machine Learning-Based Equivariant Diffusion Model Targeting the Spike Protein
The drug discovery and research for an anti-COVID-19 drug has been ongoing despite repurposed drugs in the market. Over time, these drugs were discontinued due to side effects. The search for effective drugs is still under process. The role of Machine Learning (ML) is critical in the search for nove...
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-05-01
|
| Series: | Current Issues in Molecular Biology |
| Subjects: | |
| Online Access: | https://www.mdpi.com/1467-3045/45/5/271 |
| _version_ | 1827741677783089152 |
|---|---|
| author | Vidya Niranjan Akshay Uttarkar Ananya Ramakrishnan Anagha Muralidharan Abhay Shashidhara Anushri Acharya Avila Tarani Jitendra Kumar |
| author_facet | Vidya Niranjan Akshay Uttarkar Ananya Ramakrishnan Anagha Muralidharan Abhay Shashidhara Anushri Acharya Avila Tarani Jitendra Kumar |
| author_sort | Vidya Niranjan |
| collection | DOAJ |
| description | The drug discovery and research for an anti-COVID-19 drug has been ongoing despite repurposed drugs in the market. Over time, these drugs were discontinued due to side effects. The search for effective drugs is still under process. The role of Machine Learning (ML) is critical in the search for novel drug compounds. In the current work, using the equivariant diffusion model, we built novel compounds targeting the spike protein of SARS-CoV-2. Using the ML models, 196 de novo compounds were generated which had no hits on any major chemical databases. These novel compounds fulfilled all the criteria of ADMET properties to be lead-like and drug-like compounds. Of the 196 compounds, 15 were docked with high confidence in the target. These compounds were further subjected to molecular docking, the best compound having an IUPAC name of (4aS,4bR,8aS,8bS)-4a,8a-dimethylbiphenylene-1,4,5,8(4aH,4bH,8aH,8bH)-tetraone and a binding score of −6.930 kcal/mol. The principal compound is labeled as CoECG-M1. Density Function Theory (DFT) and Quantum optimization was carried out along with the study of ADMET properties. This suggests that the compound has potential drug-like properties. The docked complex was further subjected to MD simulations, GBSA, and metadynamics simulations to gain insights into the stability of binding. The model can be in the future modified to improve the positive docking rate. |
| first_indexed | 2024-03-11T03:50:30Z |
| format | Article |
| id | doaj.art-34bff5159c13455fb5765506e9d1070d |
| institution | Directory Open Access Journal |
| issn | 1467-3037 1467-3045 |
| language | English |
| last_indexed | 2024-03-11T03:50:30Z |
| publishDate | 2023-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Current Issues in Molecular Biology |
| spelling | doaj.art-34bff5159c13455fb5765506e9d1070d2023-11-18T00:57:11ZengMDPI AGCurrent Issues in Molecular Biology1467-30371467-30452023-05-014554261428410.3390/cimb45050271De Novo Design of Anti-COVID Drugs Using Machine Learning-Based Equivariant Diffusion Model Targeting the Spike ProteinVidya Niranjan0Akshay Uttarkar1Ananya Ramakrishnan2Anagha Muralidharan3Abhay Shashidhara4Anushri Acharya5Avila Tarani6Jitendra Kumar7Department of Biotechnology, R V College of Engineering, Mysuru Road, Kengeri, Bangalore 560059, Karnataka, IndiaDepartment of Biotechnology, R V College of Engineering, Mysuru Road, Kengeri, Bangalore 560059, Karnataka, IndiaDepartment of Biotechnology, R V College of Engineering, Mysuru Road, Kengeri, Bangalore 560059, Karnataka, IndiaDepartment of Biotechnology, R V College of Engineering, Mysuru Road, Kengeri, Bangalore 560059, Karnataka, IndiaDepartment of Biotechnology, R V College of Engineering, Mysuru Road, Kengeri, Bangalore 560059, Karnataka, IndiaDepartment of Biotechnology, R V College of Engineering, Mysuru Road, Kengeri, Bangalore 560059, Karnataka, IndiaDepartment of Biotechnology, R V College of Engineering, Mysuru Road, Kengeri, Bangalore 560059, Karnataka, IndiaBangalore Bioinnovation Centre (BBC), Helix Biotech Park, Electronics City Phase 1, Bengaluru 560100, Karnataka, IndiaThe drug discovery and research for an anti-COVID-19 drug has been ongoing despite repurposed drugs in the market. Over time, these drugs were discontinued due to side effects. The search for effective drugs is still under process. The role of Machine Learning (ML) is critical in the search for novel drug compounds. In the current work, using the equivariant diffusion model, we built novel compounds targeting the spike protein of SARS-CoV-2. Using the ML models, 196 de novo compounds were generated which had no hits on any major chemical databases. These novel compounds fulfilled all the criteria of ADMET properties to be lead-like and drug-like compounds. Of the 196 compounds, 15 were docked with high confidence in the target. These compounds were further subjected to molecular docking, the best compound having an IUPAC name of (4aS,4bR,8aS,8bS)-4a,8a-dimethylbiphenylene-1,4,5,8(4aH,4bH,8aH,8bH)-tetraone and a binding score of −6.930 kcal/mol. The principal compound is labeled as CoECG-M1. Density Function Theory (DFT) and Quantum optimization was carried out along with the study of ADMET properties. This suggests that the compound has potential drug-like properties. The docked complex was further subjected to MD simulations, GBSA, and metadynamics simulations to gain insights into the stability of binding. The model can be in the future modified to improve the positive docking rate.https://www.mdpi.com/1467-3045/45/5/271equivariant diffusion modelinpaintfull-atomSARS-CoV-2molecular dynamics |
| spellingShingle | Vidya Niranjan Akshay Uttarkar Ananya Ramakrishnan Anagha Muralidharan Abhay Shashidhara Anushri Acharya Avila Tarani Jitendra Kumar De Novo Design of Anti-COVID Drugs Using Machine Learning-Based Equivariant Diffusion Model Targeting the Spike Protein Current Issues in Molecular Biology equivariant diffusion model inpaint full-atom SARS-CoV-2 molecular dynamics |
| title | De Novo Design of Anti-COVID Drugs Using Machine Learning-Based Equivariant Diffusion Model Targeting the Spike Protein |
| title_full | De Novo Design of Anti-COVID Drugs Using Machine Learning-Based Equivariant Diffusion Model Targeting the Spike Protein |
| title_fullStr | De Novo Design of Anti-COVID Drugs Using Machine Learning-Based Equivariant Diffusion Model Targeting the Spike Protein |
| title_full_unstemmed | De Novo Design of Anti-COVID Drugs Using Machine Learning-Based Equivariant Diffusion Model Targeting the Spike Protein |
| title_short | De Novo Design of Anti-COVID Drugs Using Machine Learning-Based Equivariant Diffusion Model Targeting the Spike Protein |
| title_sort | de novo design of anti covid drugs using machine learning based equivariant diffusion model targeting the spike protein |
| topic | equivariant diffusion model inpaint full-atom SARS-CoV-2 molecular dynamics |
| url | https://www.mdpi.com/1467-3045/45/5/271 |
| work_keys_str_mv | AT vidyaniranjan denovodesignofanticoviddrugsusingmachinelearningbasedequivariantdiffusionmodeltargetingthespikeprotein AT akshayuttarkar denovodesignofanticoviddrugsusingmachinelearningbasedequivariantdiffusionmodeltargetingthespikeprotein AT ananyaramakrishnan denovodesignofanticoviddrugsusingmachinelearningbasedequivariantdiffusionmodeltargetingthespikeprotein AT anaghamuralidharan denovodesignofanticoviddrugsusingmachinelearningbasedequivariantdiffusionmodeltargetingthespikeprotein AT abhayshashidhara denovodesignofanticoviddrugsusingmachinelearningbasedequivariantdiffusionmodeltargetingthespikeprotein AT anushriacharya denovodesignofanticoviddrugsusingmachinelearningbasedequivariantdiffusionmodeltargetingthespikeprotein AT avilatarani denovodesignofanticoviddrugsusingmachinelearningbasedequivariantdiffusionmodeltargetingthespikeprotein AT jitendrakumar denovodesignofanticoviddrugsusingmachinelearningbasedequivariantdiffusionmodeltargetingthespikeprotein |