In silico screening for potential COVID-19 beta-coronavirus

non-nucleoside RdRp inhibitors

Martin Shkreli1 , Kevin Mulleady2,∗ , Maureen Lohry2 , James Rondina3 , and Jason Sommer4

Abstract— COVID-19 is a nascent public health emergency as monotherapy or in combination with other anti-2019-
caused by a novel beta-coronavirus, 2019-nCoV. While industry nCoV agents such as remdesivir. We report the results of our
has made rapid and impressive progress in developing mRNA- findings, highlighting a metabolite of clofazimine as a novel,
based vaccines, neutralizing antibodies, and an intravenous
nucleoside analog RNA-dependent RNA polymerase (RdRp) well-understood candidate for the treatment of COVID-19.
inhibitor, further efforts are needed. Herein, we describe an in
silico screening campaign aimed at generating non-nucleoside II. METHODS
lead molecules with inhibitory activity against the coronavirus
RdRp protein. We reveal a metabolite of clofazimine, a widely- Using a custom HPC cluster provisioned with Autodock
used anti-leprotic and antibiotic, with global availability and Vina, we screened 130,778 compounds in parallel batches for
long-term clinical experience, as a novel RdRp inhibitor. We ability to bind to 2019-nCoV RdRp (nsp12). Our pre-selected
also find widely prescribed, FDA-approved anti-hepatitis C screening library was sourced from the ZINC15 public access
nucleoside analogs paritaprevir and ledipasvir are more potent database[7], with the majority of candidates having drug-like
inhibitors of 2019-nCoV than the investigational agent remde-
sivir. Other candidate inhibitors include bafetinib, apilimod, properties according to Lipinski’s Rule of Five.
and fimepinostat. In vitro studies are ongoing to determine
inhibitory activity, with in vivo experiments to follow. A. Protein Structure Preparation
With no available crystal or cryo-EM structure for the
I. INTRODUCTION 2019-nCoV RdRp at the time of this experiment, homology
The COVID-19 pandemic is a global health emergency modeling must be used to approximate and predict the
that requires a coordinated response across government, native, folded structure of this protein until an experimentally
academia, and industry. While impressive efforts have been derived structure becomes available. Using crystallography
made thus far, non-small molecule therapeutic modalities data from the highly homologous SARS-CoV allows the
pose significant questions. Biologic technologies such as adjustment of existing protein models for the slight, but
neutralizing antibodies are promising, but it remains to be potentially very important differences in amino acid sequence
seen if biologics manufacturing capacity will be available for and resulting protein folding of the unavailable structure.
any such therapy if proven effective. Manufacturing capacity An existing structure of the SARS-CoV RdRp is available
is traditionally a major bottleneck for the production of these as structure 6NUR in PDB[8]. With this existing structure
medicines at scale. as a template, a homology model of the 2019-nCoV RdRp
mRNA-based vaccines appear very promising, but sim- was generated using the Swiss Model web server [9], [10],
ilar manufacturing questions exist for these liposomally- [11], incorporating an NCBI protein sequence of the 2019-
encapsulated medicines; none of which have received FDA nCoV RdRp (YP_009725307.1) extracted from the 2019-
approval at the time of this writing. Furthermore, any such nCoV genome (NC_045512.2)[12] as the target sequence.
vaccine, even if proven effective, will not be useful as a In Figure 1, multiple sequence alignment demonstrates
treatment for infected coronavirus patients. that the 2019-nCoV RdRp sequence shares a 96.35% similar-
Small molecules are the bedrock of the pharmaceuti- ity with the original SARS-CoV RdRp sequence. In Figure
cal industry. Remdesivir, an RNA-dependent RNA poly- 2, the structures are also shown to have similar binding sites.
merase (RdRp) inhibitor, has in vitro and in vivo activity After validation, the protein was reduced to Chain A,
against the RdRp protein of many RNA viruses, including the location of the residues previously known to bind to
coronaviruses[1], [2], [3]. Phase III clinical trials for remde- remdesivir[13]. OpenBabel[14] was used to add additional
sivir are underway [4], [5]. Despite initial promise, remde- polar hydrogen atoms to the protein, in order to more closely
sivir results in Ebola virus patients were underwhelming [6]. resemble the environment at biological pH 7.4.
With a plethora of existing approved and investigational
nucleoside analogs for various viral proteins, we began B. Compound Dataset
screening for non-nucleotide RdRp inhibitors to be used The final compound dataset was built from various sub-
sets of the ZINC15 ligand library, including the FDA,
1 Citizen Scientist, White Deer, PA investigational-only, world-not-FDA, world, standard, in-
2 Prospero Pharmaceuticals LLC, New York, NY
3 Citizen Scientist, Fort Lee, NJ cells, in-vivo-only, and in-man-only subsets. Three sepa-
4 Citizen Scientist rate screens were performed. The first exploratory screen
∗ Corresponding Author: covid-19@prosperopharma.com (Screen 1) included all 9,524 compounds from the FDA,
 X X X X X X X X X X X
cccc ccccccccccccccccccccccccccccccccc
ACGG VV SS AA AA RR LL TT PP CCACGG TT GG TT SS TT DD VV VV YY RR AA FF DD II YY NN ED KKccccccccccccccccccccc DF
JHCC RR FF QQ EE KK DD EE ED GD NNccccLL LI DDcccccccccc ACTT MF SSccccccccccccACEE EE TT II YY NN LL VL KKccccccc
ACPP AA VV AA VK HHAcc
1 10 20 30 40 50 60 70 80 90 100
SARS-CoV
SARS-CoV-2 cccc
S
S
ADASTF
ADAQSF ccccccccccccccccccccccccccccccccc
LNRVC
LNRVC ccccccccccccccccccccc
VAGFAKFLKTNC
VAGFAKFLKTNC cccccccccc
SYFVVKRH
SYFVVKRH cccccccccccc
NYQH
NYQH ccccccc
DC
DC CccDD
X X X X X X X X X X
ccccc cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc
ACII SS RR QQ RR LL TT KK YY TT MM AA DD LL VV YY AA LL RR HHACFF DD EE GG NN CCACDD TT LL KK EE II LL VV TT YY NN CCDF
JHCC DD DD DD YY FF NN KK KK DD WWACYY DD FF VV EE NN PP DD II LL RR VV YY AA NN LL GG EE RR VV RR QQ SA LLcccccccccccc ACDD AA MM RR DN AAcc
110 120 130 140 150 160 170 180 190 200
SARS-CoV
SARS-CoV-2 ccccc
F
F
FKFRVD
FKFRID cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc
GDMVPH
GDMVPH cccccccccccc
LKTVQFC
LKTVQFC ccGG
X X X X X X X X X X
cccccccccccccccccccccc ACYY DD FF GG DD FF VI QQccVT AT PPcc
ccGG CS GGccc cccccccccccccccccc cccc ACMV DDccAT DDcc
ccLL AT KKcc
ccPP LY IIccccccccccccccc
ACDD LL LL KK YY DD FF TT EE EE RR LL CK LLccccccccccccccccc ACDD QQ TT YY HHACPP NN CCACIV NNcc
210 220 230 240 250 260 270 280 290 300
SARS-CoV
SARS-CoV-2 cccccccccccccccccccccc
I
I
VGVLTLDNQDLNGNW
VGVLTLDNQDLNGNW ccc
VPIV
VPVV cccccccccccccccccc
DSYYSLLMPILTLTRALAA
DSYYSLLMPILTLTRALTA cccc
ESH
ESH ccccccccccccccc
KW
KW ccccccccccccccccc
FDRYFKYW
FDRYFKYW
X X X X X X X X X X
ACcccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc
cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc
CLL DD DD RR CCACII LL HHDF
JHCC AA NN FF NN VV LL FF SS TT VV FF PP PP TT SS FF GG PP LL VV RR KK II FF VV DD GG VV PP FF VV VV SS TT GG YY HHACFF RR EE LL GG VV VV HHACNN QQ DD VV NN LL HHACSS SS RR LL SS FF KK EE LL LL VV YY AA AA DD PP AA MM HHACAA AA SS GG NN LL LL LL DD KK RR TT TT CCACFF SS VV AA AA
310 320 330 340 350 360 370 380 390 400
SARS-CoV C
SARS-CoV-2
X X X X X X X X X X
cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc ACFF FF FF AA QQ DD GG NN AA AA II SS DD YY DD YY YY RR YY NN LL PP TT MM CCACDD II RR QQ LL LL FF VV VV EE VV VV DD KK YY FF DD CCACYY DD GG GG CCACII NN AA NN QQ VV II VV NN NN LL DD KK
410 420 430 440 450 460 470 480 490 500
SARS-CoV
SARS-CoV-2 cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc
L
L
TNNVAFQTVKPGNFNKDFYDFAVSKGFFKEGSSVELKH
TNNVAFQTVKPGNFNKDFYDFAVSKGFFKEGSSVELKH
X X X X X X X X X X
cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc
ACGG KK AA RR LL YY YY DD SS MM SS YY EE DD QQ DD AA LL FF AA YY TT KK RR NN VV II PP TT II TT QQ MM NN LL KK YY AA II SS AA KK NN RR AA RR TT VV AA GG VV SS II CCACSS TT MM TT NN RR QQ FF HHACQQ KK LL LL KK SS II AA AA TT RR GG AA TT VV VV II GG TT SS KK FF YY GG GG WWDF
510 520 530 540 550 560 570 580 590 600
SARS-CoV
SARS-CoV-2 cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc
S
S
AGFPFNKW
AGFPFNKW JHHH NN
X X X X X X X X X X
cccccccccc ccccccccccccccccccccccccccccccc
ACLL MM GG WWACDD YY PP KK CCACDD RR AA MM PP NN MM LL RR II MM AA SS LL VV LL AA RR KK HHACNT TTccc
DF
JHCC NS LLccccccccccccccccccccccccccccccccccccccccccccccccccccc
ACRR FF YY RR LL AA NN EE CCACAA QQ VV LL SS EE MM VV MM CCACGG GG SS LL YY VV KK PP GG GG TT SS SS GG DD AA TT TT AA YY AA NN SS VV FF NN II CCACQQ AA VV
610 620 630 640 650 660 670 680 690 700
SARS-CoV
SARS-CoV-2 cccccccccc
M
M
LKTVYSDVETP
LKTVYSDVENP ccccccccccccccccccccccccccccccc
H
H ccc
C
C ccccccccccccccccccccccccccccccccccccccccccccccccccccc
SH
SH
X X X X X X X X X X
cccccccccccccccccccccccccccccccccccccc ACRR LL YY EE CCACLL YY RR NN RR DD VV DD HT ED FFcc
ccVV DN EEccccccccccccccccccccccACFF SS MM MM II LL SS DD DD AA VV VV CCACYF NNccccSS NT YYcc
ccAA AS QQccccccccccc cccccccccccccccc DF
710 720 730 740 750 760 770 780 790 800
SARS-CoV
SARS-CoV-2 cccccccccccccccccccccccccccccccccccccc
T
T
ANVNALLSTDGNKIADKYVRNLQH
ANVNALLSTDGNKIADKYVRNLQH cccccccccccccccccccccc
FYAYLRKH
FYAYLRKH ccccccccccc
GLVASIKNFKAV
GLVASIKNFKSV cccccccccccccccc
LYYQNNVFMSEAKC
LYYQNNVFMSEAKC JHWW
X X X X X X X X X X
cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc
ACEE FF CCACSS QQ HHACTT MM LL VV KK QQ GG DD DD YY VV YY LL PP YY PP DD PP SS RR II LL GG AA GG CCACFF VV DD DD II VV KK TT DD GG TT LL MM II EE RR FF VV SS LL AA II DD AA YY PP LL TT KK HHACPP NN QQ EE YY AA DD VV FF HHACLL YY LL QQ YY II RR KK LL HHACDD EE LL TT GG HHACMM LL
810 820 830 840 850 860 870 880 890 900
SARS-CoV
SARS-CoV-2 cccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccccc
T
T
ETDLTKGPH
ETDLTKGPH
X X X
cccccccccccccccccccccccccccccccc ACEE PP EE FF YY EE AA MM YY TT PP HHACTT VV LL QQ
910 920 930
SARS-CoV
SARS-CoV-2 cccccccccccccccccccccccccccccccc
D
D
MYSVMLTNDNTSRYW
MYSVMLTNDNTSRYW

Fig. 1. Multiple sequence alignment between RdRp sequences of SARS-CoV and 2019-nCoV

TABLE I
S CREEN L IGAND D ETAILS

Screen Total Subsets

1 9,524 ZINC (FDA, investigational-only, world-not-FDA,
world)
2 121,254 ZINC (standard, in-cells, in-vivo-only, in-man-
only)
3 13 RdRp-binding nucleoside candidates

C. Ligand Preparation
Before screening, GypsumDL[15] was used to convert
each compound from the SMILES format to a generated 3D
conformer with appropriate minimization and protonization.
GypsumDL was specifically chosen for the generation of fi-
nal 3D conformers because of its ability to correctly generate
Fig. 2. RdRp Chain A from SARS-CoV (magenta, PDB: 6NUR) and
2019-nCoV (cyan, homology model)
alternate tautomeric, chiral, and cis/trans isomeric forms. It
also has greater pose prediction accuracy over other utilities
such as OpenBabel Gen3D[15]. Each 3D conformer was then
converted to PDBQT with OpenBabel.
investigational-only, world-not-FDA, and world ZINC sub-
sets. The second larger screen (Screen 2) included 121,254 D. Molecular Docking and Scoring Methodology
compounds chosen from selected portions of the standard, The candidate ligands were screened against the putative
in-cells, in-vivo-only, and in-man-only ZINC subsets. Ad- binding site of remdesivir in RdRp. The center of the binding
ditional filtering was applied to Screen 2 to ensure each site was a 30Å cube centered on the location of V557 (X =
compound contained no violations of Lipinski’s Rule of Five. 141.580, Y = 149.483, Z = 164.106), a residue reported to
Because of the different compositions of subsets between the interact with remdesivir[13].
screens, Screen 1 included a much smaller, more documented AutoDock Vina was used to conduct the screen, with the
number of drug candidates than Screen 2. An additional exhaustiveness parameter set to 8, the num_modes parameter
screen (Screen 3) was prepared for investigation of several set to 9, and the energy_range parameter set to 3.
nucleoside candidates for treatment, including ribavirin, so- Both screens were run on a 512-core, cloud-based HPC
fosbuvir, galidesivir, tenofovir, paritaprevir, ledipasvir, and cluster provisioned with AutoDock Vina. Custom software
their active metabolites. Details of the ligands used for each was written to schedule an array of batched jobs, allowing
screen are displayed in Table I. multiple ligands to be screened in parallel.

2
 TABLE II TABLE III
T OP 20 C OMPOUNDS FROM S CREEN 1 (9,524 L IGANDS ) T OP 20 C OMPOUNDS FROM S CREEN 2 (121,254 L IGANDS )

ZINC ID Name kcal/mol ZINC Subset ZINC ID Name kcal/mol ZINC Subset
ZINC000095618882 clofazimine -9.9 world-not-FDA ZINC000015203333 erosnin -9.9 in-man-only
glucuronide ZINC000001070979 - -9.9 standard
ZINC000011726230 apilimod -9.9 investigational-only ZINC000000825229 - -9.9 standard
ZINC000169737127 - -9.8 investigational-only ZINC000038179173 - -9.8 in-vivo-only
ZINC000885764928 paritaprevir -9.8 world-not-FDA ZINC000000642875 - -9.8 standard
ZINC000022940637 bafetinib -9.7 investigational-only ZINC000000874254 - -9.8 standard
ZINC000139850297 - -9.7 investigational-only ZINC000001014126 - -9.8 standard
ZINC000100005073 indigo -9.7 world-not-FDA ZINC000000959864 - -9.8 standard
carmine ZINC000001130474 - -9.8 standard
ZINC000095618817 etoposide -9.7 world-not-FDA ZINC000000538290 revizinone -9.8 in-man-only
ZINC000028100198 - -9.7 investigational-only ZINC000002200016 - -9.7 standard
ZINC000073488511 fimepinostat -9.6 investigational-only ZINC000001200270 - -9.7 standard
ZINC000001530604 cromolyn -9.6 world-not-FDA ZINC000002101245 - -9.7 standard
ZINC000001539348 - -9.6 investigational-only ZINC000100005073 indigo carmine -9.7 standard
ZINC000014974132 AMG-517 -9.6 investigational-only ZINC000002299607 - -9.7 standard
ZINC000003922429 adozelesin -9.6 investigational-only ZINC000001565922 - -9.7 in-vivo-only
ZINC000003818418 R-306465 -9.6 investigational-only ZINC000006119193 12a- -9.7 in-man-only
ZINC000150338819 ledipasvir -9.6 FDA hydroxydolineone
ZINC000003915428 - -9.6 investigational-only ZINC000002301202 - -9.7 standard
ZINC000003934128 temoporfin -9.5 world-not-FDA ZINC000001956919 - -9.7 standard
ZINC000238850854 eptifibatide -9.4 world-not-FDA ZINC000103590814 - -9.7 in-cells
ZINC000206177361 endovion -9.4 investigational-only
TABLE IV
A LL C OMPOUNDS FROM S CREEN 3 (13 L IGANDS )
III. RESULTS
A. Nucleoside Analogs ZINC ID Name kcal/mol
ZINC000197964623 paritaprevir -9.6
The results of the nucleoside screen (Screen 3) are listed in ZINC000150338819 ledipasvir -9.1
Table IV. Surprisingly, our work determined anti-hepatitis C - galidesivir triphosphate -8.4
molecules such as paritaprevir and ledipasvir to have stronger - sofosbuvir triphosphate -8.3
binding affinity than remdesivir and its putative metabolites. ZINC000100074252 sofosbuvir -7.9
Hepatitis C is an RNA virus, and remdesivir, not being ZINC000012402860 ribavirin triphosphate -7.8
specifically designed for 2019-nCoV, has pan-RNA virus ZINC000013492903 galidesivir -7.5
efficacy. - remdesivir triphosphate -7.3
ZINC000013516399 tenofovir diphosphate -7.3
B. Non-Nucleosides - tenofovir monophosphate -7.2
- remdesivir -7.0
Given the plethora of nucleoside analogs, our focus with
ZINC000001035331 ribavirin -7.0
this work was on non-nucleoside ligands. Screen 1 (Table
ZINC000001543475 tenofovir -6.8
II) and Screen 2 (Table III) display ligands with the highest
binding affinity for the given binding site in RdRp. We found
clofazimine glucuronide to be one of the highest-scoring
compounds across both screens; clofazimine glucuronide vizinone as potentially novel 2019-nCoV RdRp inhibitors.
registered a score of -9.9 kcal/mol. It has a rich history of From here, many different research projects are currently in
use, specifically as an anti-infective, and a well-characterized progress, including a continuation of current work around
pharmacokinetic profile. The safety profile of clofazimine, both the computational and manual design of analogs for
however, is not fully benign[16]. Various other clinically highly scoring compounds such as clofazimine. In vitro
tested compounds also demonstrated a potent affinity for studies are ongoing to determine the inhibitory nature of
RdRp, including bafetinib, apilimod, and fimepinostat. Other these candidates, and results will soon be confirmed with
highly scoring compounds in the broad screen include eros- upcoming in vivo experiments.
nin and revizinone. Another avenue includes improvements in both virtual
The chemical structures and properties of the top-scoring screen volume and quality. Future work will include experi-
20 compounds from Screens 1 and 2 are listed in the ments on PDB 6M71, the newly released cryo-EM structure
Appendix in Tables V and VI respectively. for 2019-nCoV RdRp[17]. Screening accuracy can also be
improved with the addition of a properly chelating Mg2+
IV. CONCLUSION ion[13][18]. To process larger numbers of ligand tranches
Our results have revealed clofazimine, paritaprevir, ledi- from the ZINC dataset, GPU-accelerated virtual screening
pasvir, bafetinib, apilimod, fimepinostat, erosnin, and re- methods are being developed with consideration of different

3
knowledge-based scoring and optimization schemes, as well R EFERENCES
as previously existing improvements to the default Vina [1] Ariane J Brown, John J Won, Rachel L Graham, Kenneth H Dinnon III,
scoring function[19]. Amy C Sims, Joy Y Feng, Tomas Cihlar, Mark R Denison, Ralph S
Baric, and Timothy P Sheahan. Broad spectrum antiviral remde-
V. ACKNOWLEDGMENTS sivir inhibits human endemic and zoonotic deltacoronaviruses with
a highly divergent rna dependent rna polymerase. Antiviral research,
M. Shkreli and K. Mulleady designed and managed the 169:104541, 2019.
[2] Timothy P. Sheahan, Amy C. Sims, Sarah R. Leist, Alexandra Schäfer,
project. J. Sommer and M. Shkreli engineered the in silico John Won, Ariane J. Brown, Stephanie A. Montgomery, Alison Hogg,
workflow and performed the screening. J. Rondina provided Darius Babusis, Michael O. Clarke, Jamie E. Spahn, Laura Bauer,
high-performance computing assistance and research. M. Scott Sellers, Danielle Porter, Joy Y. Feng, Tomas Cihlar, Robert
Jordan, Mark R. Denison, and Ralph S. Baric. Comparative therapeutic
Shkreli, K. Mulleady, M. Lohry, and J. Sommer wrote the efficacy of remdesivir and combination lopinavir, ritonavir, and inter-
manuscript. feron beta against MERS-CoV. Nature Communications, 11(1):222,
December 2020.
VI. AUTHOR STATEMENT [3] Timothy P Sheahan, Amy C Sims, Rachel L Graham, Vineet D
Menachery, Lisa E Gralinski, James B Case, Sarah R Leist, Krzysztof
M. Shkreli states: Pyrc, Joy Y Feng, Iva Trantcheva, et al. Broad-spectrum antiviral
gs-5734 inhibits both epidemic and zoonotic coronaviruses. Science
The industry response to COVID-19 is inadequate. translational medicine, 9(396), 2017.
All biopharmaceutical companies should be re- [4] A Multicenter, Adaptive, Randomized Blinded Controlled Trial
sponding with all resources to combat this health of the Safety and Efficacy of Investigational Therapeutics for
the Treatment of COVID-19 in Hospitalized Adults. https:
emergency. Donations from these very valuable //clinicaltrials.gov/ct2/show/NCT04280705, last ac-
companies do not go far enough. The biopharma- cessed on 2020-04-04.
ceutical industry has a large braintrust of talent [5] Study to Evaluate the Safety and Antiviral Activity of Remdesivir
(GS-5734TM ) in Participants With Moderate Coronavirus Disease
that is not working on this problem as companies (COVID-19) Compared to Standard of Care Treatment. https:
have deprioritized or even abandoned infectious //clinicaltrials.gov/ct2/show/NCT04292730, last ac-
disease research. Medicinal chemists, structural cessed on 2020-04-04.
[6] Sabue Mulangu, Lori E Dodd, Richard T Davey Jr, Olivier Tshi-
biologists, enzymologists and assay development ani Mbaya, Michael Proschan, Daniel Mukadi, Mariano Lusak-
and research biology departments at EVERY phar- ibanza Manzo, Didier Nzolo, Antoine Tshomba Oloma, Augustin
maceutical company should be put to work until Ibanda, et al. A randomized, controlled trial of ebola virus disease
therapeutics. New England Journal of Medicine, 381(24):2293–2303,
COVID-19 is no more. 2019.
I am asking for a brief furlough (3 months) to assist [7] Teague Sterling and John J. Irwin. ZINC 15 – Ligand Discovery
for Everyone. Journal of Chemical Information and Modeling,
in research work on COVID-19. Being released to 55(11):2324–2337, November 2015.
the post-COVID world is no solace to even the [8] Robert N. Kirchdoerfer and Andrew B. Ward. Structure of the SARS-
incarcerated. As a successful two-time biopharma CoV nsp12 polymerase bound to nsp7 and nsp8 co-factors. Nature
Communications, 10(1):2342, December 2019.
entrepreneur, having purchased multiple compa- [9] Stefan Bienert, Andrew Waterhouse, Tjaart AP de Beer, Gerardo
nies, invented multiple new drug candidates, filed Tauriello, Gabriel Studer, Lorenza Bordoli, and Torsten Schwede. The
numerous INDs and clinical trial applications, I swiss-model repository—new features and functionality. Nucleic acids
research, 45(D1):D313–D319, 2017.
am one of the few executives experienced in ALL [10] Andrew Waterhouse, Martino Bertoni, Stefan Bienert, Gabriel Studer,
aspects of drug development from molecule cre- Gerardo Tauriello, Rafal Gumienny, Florian T Heer, Tjaart A P
ation and hypothesis generation, to preclinical as- de Beer, Christine Rempfer, Lorenza Bordoli, et al. Swiss-model:
homology modelling of protein structures and complexes. Nucleic
sessments and clinical trial design/target engage- acids research, 46(W1):W296–W303, 2018.
ment demonstration, and manufacturing/synthesis [11] Nicolas Guex, Manuel C Peitsch, and Torsten Schwede. Automated
and global logistics and deployment of medicines. comparative protein structure modeling with swiss-model and swiss-
pdbviewer: A historical perspective. Electrophoresis, 30(S1):S162–
For the avoidance of doubt, I have not been paid S173, 2009.
for any work on this matter or any other matter [12] National Center for Biotechnology Information (NCBI)[Internet].
while incarcerated. I do not expect to profit in Bethesda (MD): National Library of Medicine (US), National Center
for Biotechnology Information. https://www.ncbi.nlm.nih.
any way, shape or form from coronavirus-related gov/, last accessed on 2020-04-04.
treatments. I believe any company developing a [13] Leili Zhang and Ruhong Zhou. Binding mechanism of remdesivir to
coronavirus drug should seek to recoup its cost at sars-cov-2 rna dependent rna polymerase. 2020.
[14] Noel M O’Boyle, Michael Banck, Craig A James, Chris Morley, Tim
most and be willing to perform the work as a civil Vandermeersch, and Geoffrey R Hutchison. Open babel: An open
service at the least. If the government is willing to chemical toolbox. Journal of cheminformatics, 3(1):33, 2011.
reward industry for their work on this catastrophic [15] Patrick J Ropp, Jacob O Spiegel, Jennifer L Walker, Harrison Green,
Guillermo A Morales, Katherine A Milliken, John J Ringe, and
situation, it will be at each company’s discretion to Jacob D Durrant. Gypsum-dl: an open-source program for preparing
accept, negotiate or deny such funding, including small-molecule libraries for structure-based virtual screening. Journal
bulk purchases, cost reimbursement, tax credits and of cheminformatics, 11(1):34, 2019.
[16] Mack R Holdiness. Clinical pharmacokinetics of clofazimine. Clinical
other benefits. pharmacokinetics, 16(2):74–85, 1989.
Because of temporary communication restrictions aimed at [17] Yan Gao, Liming Yan, Yucen Huang, Fengjiang Liu, Yao Zhao, Lin
Cao, Tao Wang, Qianqian Sun, Zhenhua Ming, Lianqi Zhang, et al.
curtailing the spread of COVID-19, the author, M. Shkreli, Structure of rna-dependent rna polymerase from 2019-ncov, a major
was unable to have final review over this work. antiviral drug target. bioRxiv, 2020.

4
[18] Jacopo Sgrignani and Alessandra Magistrato. The structural role
of mg2+ ions in a class i rna polymerase ribozyme: a molecular
simulation study. The Journal of Physical Chemistry B, 116(7):2259–
2268, 2012.
[19] Rodrigo Quiroga and Marcos A Villarreal. Vinardo: A scoring
function based on autodock vina improves scoring, docking, and
virtual screening. PloS one, 11(5), 2016.

5
 VII. APPENDIX
The numbers of hydrogen bond donors and acceptors for a single ligand are displayed in the HBD and HBA columns
respectively.

TABLE V: Top 20 ZINC compounds from Screen 1 (9,524 ligands)

Score MW
Structure ZINC ID Name logP HBD HBA Subset
(kcal/mol) (g/mol)

clofazimine world-not-
ZINC000095618882 -9.9 665.6 2.8 5 11
glucuronide FDA

investigational-
ZINC000011726230 apilimod -9.9 418.5 4.6 1 8
only

investigational-
ZINC000169737127 - -9.8 800.9 0.1 9 10
only

world-not-
ZINC000885764928 paritaprevir -9.8 765.9 4.7 3 10
FDA

investigational-
ZINC000022940637 bafetinib -9.7 576.6 4.2 2 11
only

investigational-
ZINC000139850297 - -9.7 571.4 6.8 2 12
only

indigo world-not-
ZINC000100005073 -9.7 466.4 -3.6 2 9
carmine FDA

Continued on next page

6
 TABLE V: Top 20 ZINC compounds from Screen 1 (9,524 ligands)

Score MW
Structure ZINC ID Name logP HBD HBA Subset
(kcal/mol) (g/mol)

world-not-
ZINC000095618817 etoposide -9.7 588.6 1.1 3 13
FDA

investigational-
ZINC000028100198 - -9.7 381.4 4.0 1 5
only

investigational-
ZINC000073488511 fimepinostat -9.6 508.6 1.3 2 12
only

world-not-
ZINC000001530604 cromolyn -9.6 468.4 1.9 3 11
FDA

investigational-
ZINC000001539348 - -9.6 476.5 5.6 2 4
only

investigational-
ZINC000014974132 AMG-517 -9.6 430.4 4.6 1 9
only

investigational-
ZINC000003922429 adozelesin -9.6 502.5 4.4 3 4
only

investigational-
ZINC000003818418 R-306465 -9.6 413.5 1.3 2 8
only

ZINC000150338819 ledipasvir -9.6 889.0 7.4 4 10 FDA

investigational-
ZINC000003915428 - -9.6 567.5 -0.8 6 13
only

Continued on next page

7
 TABLE V: Top 20 ZINC compounds from Screen 1 (9,524 ligands)

Score MW
Structure ZINC ID Name logP HBD HBA Subset
(kcal/mol) (g/mol)

world-not-
ZINC000003934128 temoporfin -9.5 680.7 8.8 6 6
FDA

world-not-
ZINC000238850854 eptifibatide -9.4 832.0 -2.4 10 12
FDA

investigational-
ZINC000206177361 endovion -9.4 495.2 4.2 3 10
only

8
 TABLE VI: Top 20 ZINC compounds from Screen 2 (121,254 Ligands)

Score MW
Structure ZINC ID Name logP HBD HBA Subset
(kcal/mol) (g/mol)

ZINC000015203333 erosnin -9.9 320.0 4.1 0 5 in-man-only

ZINC000001070979 - -9.9 497.1 4.2 1 6 standard

ZINC000000825229 - -9.9 459.1 6.3 1 5 standard

ZINC000038179173 - -9.8 466.1 5.4 2 6 in-vivo-only

ZINC000000642875 - -9.8 441.2 4.0 2 6 standard

ZINC000000874254 - -9.8 491.1 4.2 2 5 standard

ZINC000001014126 - -9.8 481.1 3.2 2 3 standard

ZINC000000959864 - -9.8 449.1 5.9 3 7 standard

Continued on next page

9
 TABLE VI: Top 20 ZINC compounds from Screen 2 (121,254 Ligands)

Score MW
Structure ZINC ID Name logP HBD HBA Subset
(kcal/mol) (g/mol)

ZINC000001130474 - -9.8 370.1 4.0 1 5 standard

ZINC000000538290 revizinone -9.8 459.2 3.7 1 6 in-man-only

ZINC000002200016 - -9.7 428.0 2.3 3 8 standard

ZINC000001200270 - -9.7 433.0 2.6 2 5 standard

ZINC000002101245 - -9.7 466.2 4.9 0 4 standard

indigo
ZINC000100005073 -9.7 422.0 0.2 4 8 standard
carmine

ZINC000002299607 - -9.7 416.1 0.9 2 8 standard

ZINC000001565922 - -9.7 430.1 6.1 4 6 in-vivo-only

12a-
ZINC000006119193 hydroxy- -9.7 352.1 2.8 1 7 in-man-only
dolineone

ZINC000002301202 - -9.7 429.2 2.9 1 7 standard

Continued on next page

10
 TABLE VI: Top 20 ZINC compounds from Screen 2 (121,254 Ligands)

Score MW
Structure ZINC ID Name logP HBD HBA Subset
(kcal/mol) (g/mol)

ZINC000001956919 - -9.7 418.1 2.9 3 6 standard

ZINC000103590814 - -9.7 466.2 5.2 1 9 in-cells

11