Identification of good and bad fragments of tricyclic triazinone analogues as potential PKC-θ inhibitors through SMILES–based QSAR and molecular docking

TLDR: The developed QSAR models are descriptive and predictive in nature and the results are in sound agreement with the experimental observations.

Abstract: Based on the mechanism of action of PKC-θ, the inhibition of this enzyme is considered a potential target for the treatment of autoimmune diseases such as rheumatoid arthritis (RA), inflammatory bowel disease (IBD), and psoriasis. In the present study, 57 structurally diverse tricyclic triazinone analogues as potential PKC-θ inhibitors has been taken into consideration for QSAR analysis through Monte Carlo optimization. QSAR models are developed using the balance of correlation method in the CORAL software with two target functions (TF1 and TF2). The models constructed with IIC are found more robust and authentic. The established QSAR model with best $$ {R}_{\mathrm{calibration}}^2 $$  = 0.9653 for split 3 is considered the topmost model. The predictabilities of the recommended QSAR model are assessed through various statistical parameters. The outlier of each model is also identified using the applicability domain (AD). The common mechanistic interpretation of the increasing and decreasing attributes has been extracted by evaluating the correlation weights of diverse structural attributes obtained in three Monte Carlo optimization runs from two splits, i.e., split 3 and 4. In the last, the result of mechanistic interpretation is validated by performing the docking studies of compounds PKC03, PKC07, PKC16, PKC25, and PKC56 in the experimental structure of protein kinase C-θ (PDB ID: 4Q9Z). The numerical value of the correlation coefficient between calculated activity and binding affinity is found 0.9597. Hence, the developed QSAR models are descriptive and predictive in nature and the results are in sound agreement with the experimental observations.