CACHE Challenge #2: Targeting the RNA Site of the SARS-CoV-2 Helicase Nsp13

Abstract

A critical assessment of computational hit-finding experiments (CACHE) challenge was conducted to predict ligands for the SARS-CoV-2 Nsp13 helicase RNA binding site, a highly conserved COVID-19 target. Twenty-three participating teams comprised of computational chemists and data scientists used protein structure and data from fragment-screening paired with advanced computational and machine learning methods to each predict up to 100 inhibitory ligands. Across all teams, 1957 compounds were predicted and were subsequently procured from commercial catalogs for biophysical assays. Of these compounds, 0.7% were confirmed to bind to Nsp13 in a surface plasmon resonance assay. The six best-performing computational workflows used fragment growing, active learning, or conventional virtual screening with and without complementary deep-learning scoring functions. Follow-up functional assays resulted in identification of two compound scaffolds that bound Nsp13 with a K_d below 10 μM and inhibited in vitro helicase activity. Overall, CACHE #2 participants were successful in identifying hit compound scaffolds targeting Nsp13, a central component of the coronavirus replication-transcription complex. Computational design strategies recurrently successful across the first two CACHE challenges include linking or growing docked or crystallized fragments and docking small and diverse libraries to train ultrafast machine-learning models. The CACHE #2 competition reveals how crowd-sourcing ligand prediction efforts using a distinct array of approaches followed with critical biophysical assays can result in novel lead compounds to advance drug discovery efforts.

Figure 1:. Fragments occupy the SARS-CoV-2 Nsp13 RNA-binding channel.

Composite image formed by superimposing experimental structures of Nsp13 in complex with four fragments and in complex with RNA and ADP (blue and orange respectively; PDB code 7RDY). CACHE #2 participants were asked to find ligands targeting the RNA-binding site occupied by fragments. Electrostatic potential coloring of the binding site, revealing the overall polar area, and bound fragments are depicted in the inset.

Figure 2:. Computational workflows used in CACHE #2

Figure 3:. Drug-likeness and chemical diversity of 1957 Round 1 compounds.

(a) Number of compounds tested in Round 1 and advanced to Round 2 for each participant. (b) Chemical descriptors distribution of Round 1 compounds. (c) Pairwise Tanimoto distance matrix, using ECFP4 Morgan fingerprints from RDKit (compounds are ordered based on the selected teams). (d) closest analogs selected by different participants. MW: molecular weight; PSA: polar surface area; HBD: hydrogen-bond donors; HBA: hydrogen-bond acceptors; ROTB: rotatable bonds; TD: Tanimoto Distance.

Figure 4:. Experimental evaluation of CACHE #1 Round 1 compounds.

Binding to Nsp13 measured by SPR and ATPase activity inhibition was used to advance compounds to Round 2.

Figure 5:. Top six chemical series identified in Round 2.

Activity of the parent molecules and experimental data from Round 2 analogs are shown, including SPR sensorgrams and ¹⁹F-NMR spectra. Computational workflow IDs are encoded into compound names.

Figure 5:. Top six chemical series identified in Round 2.

Activity of the parent molecules and experimental data from Round 2 analogs are shown, including SPR sensorgrams and ¹⁹F-NMR spectra. Computational workflow IDs are encoded into compound names.

Figure 5:. Top six chemical series identified in Round 2.

Activity of the parent molecules and experimental data from Round 2 analogs are shown, including SPR sensorgrams and ¹⁹F-NMR spectra. Computational workflow IDs are encoded into compound names.

Figure 6:. Two compounds inhibited RNA duplex unwinding.

Out of the 13 most potent compounds in the SPR assay (Table S7), CACHE2-HO-1431_6 and CACHE2-HO_1454_15 had measurable IC₅₀ values in a FRET-based RNA unwinding assay (a),- and had a detectable inhibitory effect in a gel-based RNA unwinding assay when added at 1 mM (b).

Figure 7:. Scores of CACHE2 participants.

For each team, the aggregated score of all Round 1 hits (a) or the score of the best Round 1 hit (b) selected from the Enamine Real library is plotted. (c) Normalized score when predicting active molecules from the 1957 Round 1 compounds (calculated as the aggregated score of all compounds predicted active divided by the number of compounds predicted active). The score of each molecule was assigned by the CACHE Hit Evaluation Committee (Table S1).

Figure 8:. Best performing workflows.

(a) Group, workflow ID, and associated ranks in three evaluation schemes. (b) Schematics of the computational workflows.

Figure 9:. Docked poses of top compounds.

The docked poses of some of the best scoring CACHE #2 hits (right) compared with the crystal structure of fragments found in the PDB (left). RNA from a superimposed cryo-EM structure shown in blue (PDB code: 7RDY). CACHE_1454_98 was docked to an alternate site.

Figure 10:. Classification of most successful workflows.

Computational workflows are classified based on hit-prediction strategies. Computational steps using machine-learning are highlighted in blue. Software names are shown in italic.