Galaxy workflows for fragment-based virtual screening: a case study on the SARS-CoV-2 main protease

Abstract

We present several workflows for protein-ligand docking and free energy calculation for use in the workflow management system Galaxy. The workflows are composed of several widely used open-source tools, including rDock and GROMACS, and can be executed on public infrastructure using either Galaxy's graphical interface or the command line. We demonstrate the utility of the workflows by running a high-throughput virtual screening of around 50000 compounds against the SARS-CoV-2 main protease, a system which has been the subject of intense study in the last year.

Keywords: Computational chemistry; Fragment screening; SARS-CoV-2; Workflows.

Fig. 1

Schematic of the docking and scoring workflow

Fig. 2

Schematic of the MMGBSA workflow. A modular subworkflow for system parameterization is shared with the dcTMD workflow; see Fig. 4 for details

Fig. 3

Pull groups for the TMD simulations (image depicts the x0397 structure). Group 1 (cyan) consists of the ligand non-hydrogen atoms. Group 2 (green) consists of a selection of alpha-carbons in the Mpro active site. During the course of the TMD simulation, the two groups are pulled apart by means of a constant constraint force

Fig. 4

Schematic of the dcTMD workflow

Fig. 5

a and b Distributions of SuCOS and TransFS scores per fragment; the mean values are marked in black. c Scatter plot of SuCOS and TransFS scores for all poses. 209 of these are filtered for further screening d All fragments superimposed on the protein structure and colored by the main subpocket to which the fragment binds (S1’ red, S1 blue, S2 pink, S3 orange)

Fig. 6

Plot of MMGBSA enthalpies for poses derived from each of the 22 fragments (mean marked by the large circles)

Fig. 7

Ligands (cyan) binding in pockets, overlaid on the parent fragments (green): S1’ a) (x0397; SuCOS 0.65)), S1 b (x0387; SuCOS 0.56), S2 c (x0678; SuCOS 0.53) and S3 d (x0161; SuCOS 0.60)

Fig. 8

Free energy curves derived from dcTMD calculations for two of the screened compounds

Fig. 9

a Friction profiles for four selected ligands; the profiles for the ligands binding in subpocket S1/S1’ (red/pink) show a rise starting at 0.2 nm, whereas for those binding in subpocket 2 (blue/cyan), this is absent, with an increase being observable instead at 0.3 nm. b Ligands exiting the subpocket S1/S1’ at 0.25 nm from the initial binding position (pink), with Asn142 highlighted, and subpocket S2 at 0.33 nm from the initial binding position (green), with Ser46 highlighted

Fig. 10

The average number of interactions observed and the free energy as calculated by MMGBSA are correlated ($R^2$ = − 0.46). The weakness of the relationship reflects the high variation in the strength and importance of interactions

Fig. 11

Maximum dcTMD free energy scores for compounds which display hydrogen bonding with the peptide backbone at residues Gly143 ($R^2=0.69$) and Cys145 ($R^2=0.85$)