Context: Mangosteen (Garcinia mangostana L.) is used in traditional medicine as an antibacterial, antioxidant, and anti-inflammatory.

Aims: To determine the molecular mechanism and potential of garciniaxanthone derivate compounds from G. mangostana as SARS-CoV-2 antiviral and prevent cytokine storm through in silico approach.

Methods: Ligand and protein samples were obtained from databases such as PubChem and Protein Databank, then drug-likeness analysis using Lipinski, Ghose, Veber, Egan, and Muege rules on SwissADME server, prediction of antiviral probability through PASSOnline server. Furthermore, molecular docking simulation with PyRx v1. 0 software (Scripps Research, USA) with an academic license, identification of interactions and chemical bond positions of ligands on the target by PoseView server, 3D visualization of PyMOLv. 2.5. 2 software (Schrödinger, Inc., USA) with an academic license, molecular dynamics simulation for molecular stability prediction by CABS-flex v2. 0 server, target prediction of antiviral candidate compounds by SwissTargetPrediction server, pathway analysis through STRING v11. 5 database, and toxicity by ProTox-II server were used.

Results: Garciniaxanthone C from G. mangostana was found to be a drug-like molecule with low toxicity. This can be a candidate for SARS-Cov-2 antiviral through inhibitor activity on two viral enzymes consisting of Mpro and replicase with a binding affinity value that is more negative than other garciniaxanthone derivates and is stable. Garciniaxanthone C is predicted to bind and inhibit pro-inflammatory proteins that trigger cytokine storms, such as NFKB1 and PTGS2.

Conclusions: Garciniaxanthone derivative compounds from G. mangostana may be candidates for SARS-CoV-2 antiviral and preventing cytokine storm through garciniaxanthone C activity.