Predicting inhibitors for SARS-CoV-2 RNA-dependent RNA polymerase using machine learning Nazim Medzhidov, PhD Elix, Inc. 2020.10.28

Outline 2 ● Introduction & Target Selection ● Data Collection ● Models ● Results & Discussion

Target Selection ● RNA dependent RNA polymerase (RdRp): ○ Present in almost all RNA viruses ○ “Right hand” with Thumb, finger and palm subdomains ○ Core structural similarity between RdRps of different RNA viruses & conserved amino acids in the active site (palm subdomain) ○ Motifs A-G ○ Structural similarity to HCV and poliovirus RdRps 3 Gao Y, Yan L, Huang Y, et al. Structure of the RNA-dependent RNA polymerase from COVID-19 virus [published online ahead of print, 2020 Apr 10]. Science. 2020;eabb7498. doi:10.1126/science.abb7498 Gao Y, et al., Science, 2020

Data Collection 4 ● PubChem and ChEMBL Bioassays ● RdRp inhibitors with measured activity values (IC50/EC50) ● Training/validation set (1356 entries: 700 = active, 656 = inactive) ● Test set: 20 pre-clinical RdRp inhibitors and 20 kinase inhibitors. ● FDA approved drugs and drugs at the clinical stage: ■ Antiviral ■ Anti-inflammatory

Data Collection 5 ● Main targets in the training/validation set: ○ HCV RdRp: (+ssRNA, Flaviviridae) [1202 entries: 676 = active, 526 = inactive] ○ Dengue virus RdRp: (+ssRNA, Flaviviridae) [67 entries: 15 = active, 52 = inactive] ○ Poliovirus RdRp: (+ssRNA, Picornaviridae) [24 entries, 0 = active 24 = inactive] ○ Influenza virus RdRp: (-ssRNA, Orthomyxoviridae) [45 entries: 5 = active, 40 = inactive] ○ Others: (+ssRNA & -ssRNA) [18 entries: 4 = active, 14 = inactive]

Results 6

Results: Best models performance on the test set TP: True Positives TN: True Negatives FP: False Positives FN: False Negatives 7

Ensemble Model Results 8 TP: True Positives TN: True Negatives FP: False Positives FN: False Negatives

Antiviral Drugs 9 Beclabuvir: HCV RdRp, non-nucleoside inhibitor Asunaprevir, Paritaprevir, Faldaprevir: HCV protease

Anti-inflammatory Drugs 10 Betulinic acid and Ursolic acid: naturally occuring drugs with reported anti-HIV-1 activity

Summary 11 ● RdRp inhibitors dataset with experimental activity values was established ● Results of machine learning models compared to Autodock Vina predictions ● Remdesivir was predicted as a potential drug candidate ● Several HCV protease inhibitors were identified as potential candidates. ● Some predicted drug candidates had low binding energy scores against SARS-CoV-2 RdRp (as calculated by Autodock Vina) Cozac, R., Medzhidov, N., Yuki, S. (2020). Predicting inhibitors for SARS-CoV-2 RNA-dependent RNA polymerase using machine learning and virtual screening https://arxiv.org/abs/2006.06523

/in/nazim-med/ 12 Questions & Contact For other questions and further discussions, please reach out on LinkedIn. Thanks for joining!

株式会社Elix http://ja.elix-inc.com/ 1