A metabolic modeling approach reveals promising therapeutic targets and antiviral drugs to combat COVID-19

Abstract

In this study we have developed a method based on Flux Balance Analysis to identify human metabolic enzymes which can be targeted for therapeutic intervention against COVID-19. A literature search was carried out in order to identify suitable inhibitors of these enzymes, which were confirmed by docking calculations. In total, 10 targets and 12 bioactive molecules have been predicted. Among the most promising molecules we identified Triacsin C, which inhibits ACSL3, and which has been shown to be very effective against different viruses, including positive-sense single-stranded RNA viruses. Similarly, we also identified the drug Celgosivir, which has been successfully tested in cells infected with different types of viruses such as Dengue, Zika, Hepatitis C and Influenza. Finally, other drugs targeting enzymes of lipid metabolism, carbohydrate metabolism or protein palmitoylation (such as Propylthiouracil, 2-Bromopalmitate, Lipofermata, Tunicamycin, Benzyl Isothiocyanate, Tipifarnib and Lonafarnib) are also proposed.

Figure 1

Predicted relative effects of blocking metabolic enzymes, on the healthy lung cells and the infected cells, respectively (A). Schema summarizing the role of the identified drug targets on the assembly of virions (B). Illustration of how the inhibition of an enzyme modifies the space of feasible metabolic flux distributions. The method used in this work searches to decrease the metabolic capabilities for the formation of virions, while keeping limited effects on the formation of cellular biomass (C).

Figure 2

Co-factor binding site of FAR2 with the binding positions of NADPH (green) lonafarnib (red) and tipifarnib (yellow). Both drugs interact with the co-factor binding site (A). Fatty acyl-CoA reductase (FAR2) bound to NADPH (B). Docking revealed a binding affinity of − 9.5 kcal/mol for FAR2 and NADPH; one of its natural substrates (C), while the predicted binding affinities of Lonafarnib (D) and Tipifarnib (E) are − 10.9 kcal/mol and − 9.5 kcal/mol, respectively. This suggests that these two drugs (with a preference for Lonafarnib) could displace NADPH from its binding site and block the activity of FAR2.

Figure 3

Structure of the CYB5R3 binding site of Propylthiouracil (A), with a binding ΔG of − 5.4 kcal/mol (B) and docking of ZINC39395747 (ΔG =  − 7.5 kcal/mol) (C) and ZINC05626394 (ΔG =  − 6.5 kcal/mol) (D).