Design of In-Silico Inhibitor COX-2 using DEEPScreen-QSAR and Molecular Docking Sudarko, Nahdiatul Ummah and Anak Agung Istri Ratnadewi
Department of Chemistry, Faculty of Mathematics and Natural Science, University of Jember, Jl. Kalimantan 37, Jember, 68121, East Java, Indonesia
Abstract
HCC is cancer that develops from abnormal cell growth in the liver. Abnormal cell growth causes persistent inflammation. The protein that plays a role in cancer inflammation is COX-2 with the activation site located in the amino acids Tyr385 and Ser530. HCC can be prevented by blocking the active site of COX-2 using drugs (inhibitors). Machine learning will screen thousands to millions of compounds that have potential as anti-inflammatory HCC drugs by virtual screening. The classification-based virtual screening used is DEEPScreen-QSAR modeling. The Machine learning has been trained with dataset obtained from previous research to get the best QSAR model. The best QSAR model obtained has an accuracy of 0.776. The model was used for the virtual screening of 1,914,538 small molecules. The results of virtual screening using the DEEPScreen-QSAR model obtained were 796,810 molecules being active. Active molecules from the DEEPScreen virtual screening need to be re-screened using Lipinski RO5. The screening results obtained as many as 580,881 active molecules with 0 deviations (not violating Lipinski^s RO5), then the activity value was predicted. Prediction of the activity value/pIC50 of the drug candidate was carried out using QSAR regression. Evaluation of the regression QSAR model resulted in a comparison curve between the predicted value of activity/pIC50 and the value of activity/pIC50 experimentally in the form of R=0.694- R2=0.4818- MSE=5,253 and RMSE=2,291. The stage after the virtual screening is then validated or scored with molecular docking using AutodockVina in PyRx to determine the Gibbs^s free energy. Then visualization was carried out in 3D and 2D using PyMOL and BDS to determine the interaction between the ligand (potential drugs) and protein (COX-2). Based on the results of ligand and structure-based screening, recommendations for anti-inflammatory drugs in HCC were obtained, namely molecules with the code CHEMBL372052, CHEMBL366063, CHEMBL3941655, and CH
