In spite of the previously known coronavirus outbreaks in 2002 and 2003 followed later in 2012 with MERS-CoV spread in middle-Asia, no effective treatment has been developed. Shortage of financial support and attention in R&D area resulted in a desperate outcome. The world appeared to be largely unprepared for the new outbreak, caused by SARS-CoV-2, which began in December 2019 and quickly grew into a global pandemic. The past research on SARS/MERS-CoV family of viruses has not produce any lead series which can be quickly progressed into marketed drugs.

An unprecedented biotech race is now on to design cures for COVID-19 and to curb the crisis until a reliable vaccine is available and widely adopted. While most of the efforts are focused on drug repurposing strategies, de-novo drug discovery is essential for achieving sufficient specificity and efficacy towards the new SARS-CoV-2 virus.

To meet the urgent need for potent new antivirals against COVID-19, Enamine has designed and preplated for quick delivery a small-molecule compound library focusing on SARS-CoV most promising targets. This library incarnates Enamine’s rich experience in design and synthesis of antiviral compounds. We take an active part in the global open science initiative COVID Moonshot project, providing synthesis, ADME-Tox profiling, medicinal chemistry and logistic services.

The Coronavirus Library consists of 7 sublibraries which can be also acquired separately.

Library design

To design our library we carefully collected all available structural information of most promising SARS-CoV protein targets. The protein structures were scrutinized and docking models were created. The docking models were validated by short MD simulations and verified then by ability to form complexes with reported active molecules. We have focused on the following 5 targets and NSP, reported to be most important for virus transmission and essential for viral replication:

• SARS-CoV-2 main protease M^pro (also called 3CL^pro)
• RNA-dependent RNA polymerase (RdRp)
• Papain-like protease (PLpro)
• Angiotensin-converting enzyme 2 (ACE2) receptor
• Type-II transmembrane serine protease (TMPRSS2)
• Non-structual proteins (NSP) of SARS-CoV-2, with reported 3D-struture.

For cysteine and serine proteases (M^pro and TMPRSS2) covalent docking has been carried out to identify promising covalent binders, which can elongate inhibition action. The databases of screening compounds were preliminary filtered to contain only compounds with specific warheads that are selective to each amino acid:

All other target-focused sublibraries passed stringent MedChem filters, including PAINS and Molecular Parameters restrictions to represent mainly lead-like space. The hits derived from our library can be easily followed with analogs and growing strategy to maximize MedChem variability within the structure.

Examples of the docking results

Binding poses of two predicted hit molecules after docking calculation on M^pro protease Z1609752806 (light blue ligand on the left side) and Z1143050660 (grey ligand on the right side).

Reliable structural data of a target protein is the key to a successful drug discovery program. This knowledge is extremely important for the efficient development of selective and potent small molecule drugs. Recent achievements in the field of protein structure investigation are impressive and have enabled development of new approaches and screening techniques. Combination of modern, elaborated crystallographic methods with smart-design of chemical libraries can make a breakthrough in our understanding of protein structure changes and behavior.

Novel crystallographic screening methodology reported by O’Reilly et al. in Drug Discovery Today 2019 was developed at Astex Pharmaceuticals, Cambridge, UK. The high-concentration aqueous soaks were made with a chemically diverse and ultra-low-molecular-weight fragment library “MiniFrags” (heavy atom count 5–7). This allowed identification of hot and warm spots on proteins. High screening hit rates reflect enhanced sampling of chemical space. MiniFrag screening can represent thus a highly effective method for guiding optimisation of fragment-derived lead compounds.

We have collaborated with the Astex’ scientists on making MiniFrag library available to the research community in the most convenient format. Screening at 1M suggests that the fragments would be provided dry, ready for dissolution prior to protein soaking.

Library design

The compounds in the MiniFrag Library offered by Enamine are identical to those proposed by Astex. Their structures and selection principles are described in O’Reilly et al. in Drug Discovery Today, Volume 24, Issue 5, May 2019, Pages 1081-1086. DOI: 10.1016/j.drudis.2019.03.009.

Fragment screening is an efficient way of establishing initial points for drug discovery. However, seemingly simple small molecules don’t necessarily mean that their chemistry is not complex. Quite the contrary, fragment hits may be challenging to follow up. Synthesis routes can be long and rigid not allowing making small specific changes while growing, linking or merging the hit.

It was shown by Cox et al., in Chem. Sci. 2016 that the compounds synthesised from a robust and general synthetic reaction can be prioritized for building fragment library. Identification of so-called “poised” bonds in a fragment means that a library of analogues can be rapidly elaborated using standard parallel chemistry. First poised library was developed at Diamond Light Source, UK (Diamond) and Structural Genomic Consortium, UK (SGC) in the frame of one of the iNEXT Joint Research Activities. Enamine collaborated with the joint research alliance in design of a next generation of DSI-poised Library. DSI stands for Diamond, SGC and iNEXT. On January 10 2018 Diamond and SGC announced that Enamine would become a key supplier of poised fragment and analogue libraries to the XChem Facility in Oxford, UK.

Library design

The original library developed by Diamond and SGC has been enhanced with a careful selection of fragments synthesized at Enamine via parallel chemistry from its building blocks, which are widely commercially available. Every researcher will be available to quickly synthesize analogues of the poised fragments because such building blocks are not complex and are exemplified with many analogues on the market. The reactions are frequently used in parallel chemistry: amide coupling, reductive amination, Suzuki-Miyaura coupling, aromatic nucleophilic substitution, synthesis of sulfonamides, ureas, etc.

Examples of the molecules

The library was designed to find molecules which target both end-chains of water channels (Aquaporins) AQ1, AQ4, AQ5 monomers and internal central pore of AQ5, that possess same physico-chemical properties as those in AQ1 and AQ4. A small number of known active inhibitors against this type of targets do not allowed us to build qualitative QSAR model or pharmacophore models. The HTS docking approach was used to search of new potential inhibitors, 2D similarity search was carried out to enrich the library with compounds similar to reported actives.

Library Design

Docking optimization & calculations

Three protein structures (3D9S, 1IH5, 3GD8) were chosen after analysis of 11 available X-Ray structures (1H6I, 1IH5, 2ZZ9, 3GD8, 4NEF, 3D9S, 5DYE, etc.). However, it is known that shift in open/closed state is dramatically sensitive to some mutations. This data described several hot-spots, which are located inside the channel and were used for preparation of docking models on the primary stage of protein structure refinement, of course if these residues are involved in interaction, not in rigidity of the helix. Additionally the whole surface of each monomer was probed with FT-MAP server, which showed itself as a very accurate and trustworthy tool for binding site identification. It is based on evaluation of energy interaction between small organic probes (solvents, saturated and unsaturated cycles...) and the surface of interest.

The design of aquaporin library was performed with implementation of two different approaches: flexible docking and high throughput screening. Flexible docking protocol assumed a partially-free rotatable motion of side-chains, which are located in the entrance of both cavities in AQ5/AQ4/AQ1 proteins. The area of flexibility covered about 5 Ų around the centre of the channel in one unit of the tetrameric structure. The centre of a Grid box (determination of the binding area) was selected based on residue mutation studies and FT-MAP calculations, mentioned above. To expand the volume of putative binding sites a set of real inhibitors, which were disclosed in publications and chemblDB, were docked together with compounds, which were shown to interact with the target (AqB013, Bumetanide, Zonisamide, Arbidol and Acetazolamide were taken from the literature and some neurotransmitters, like dopamine, epinephrine and serotonine, were chosen based on our collaboration with academicians.

In this way a more suitable conformation of the gate was achieved in binding sites and these optimized structures were applied for screening procedure with slightly less exhaustive parameters of pose estimation and number of generated conformers. Here it is a short example of parametrization if the docking into the AQ5 model. Among constraints, which formed the filtering sieve, a hydrophobic core inside the channel and partial match against the set of amino acids (ARG:188, ASN:120, GLY:181, A:HIS:173, VAL:119 for AQ5 site2; and GLN:81, HIE:67, PRO:62, SER:152, THR:150, THR:150, TYR:90, TYR:90 for AQ5 site1), should be mentioned. All of them are shown in the figure. Those compounds which possessed a hydrophobic moiety and satisfied another condition (2-3 h-bond interactions with any of the listed residues), reached the next stage of more accurate and time-consuming docking procedure. Rating of these compounds is a combination of empiric values, post-processing calculation of energy, steric and h-bonding impacts. The more higher rating, the more probable binding of the ligand.