<![CDATA[Background: Recurrent aphthous stomatitis (RAS) is a common ulcerative disorder of the oral mucosa in which cyclooxygenase-2 (COX-2)-derived prostaglandin E2 contributes to pain and inflammation. Stem sap of Musa paradisiaca var. Sapientum (Ambonese banana) is used ethnomedicinally for oral ulcers, but the constituents responsible for its anti-inflammatory effect are unidentified. Objective: To screen phytochemicals previously reported from M. paradisiaca stem sap for predicted binding to human COX-2 and to characterise their predicted pharmacokinetic (ADMET) profiles, in order to prioritise candidates for experimental testing. Methods: Nine phytochemicals compiled from published phytochemical reports, together with celecoxib and the co-crystallised ligand rofecoxib, were docked into human COX-2 (PDB ID 5KIR). The protocol was validated by redocking rofecoxib. ADME-T descriptors were predicted with SwissADME and ADMETlab 3.0. Result: The docking protocol was successfully validated (RMSD = 0.992 Å). The five best-scoring compounds were catechin (docking score = −8.76 kcal/mol; predicted Ki = 376.01 nM), gallocatechin (−8.59; 504.97 nM), quercetin-3-O-glucoside (−8.43; 658.19 nM), quercetin (−8.38; 721.52 nM) and leucocyanidin (−8.37; 734.98 nM). Each formed hydrogen bonds with Tyr385, Ser530 or Gln192 and hydrophobic contacts with Leu352 and Val523. All five were predicted to have low Caco-2 permeability and low human intestinal absorption, and four carried a high predicted skin-sensitisation probability. Conclusion: Catechin and gallocatechin were the best-scoring candidate COX-2 inhibitors. Their binding affinity and favorable topical pharmacokinetics support their potential as non-steroidal anti-inflammatory agents for RAS. Further evaluation of COX-1/COX-2 selectivity, in vitro, and in vivo experimental validation are required to confirm these findings.]]>
