Various in silico approaches were utilized to design quinazoline derivatives as anticancer agents targeting VEGFR-2. Six 2D-QSAR models were generated via Monte Carlo optimization method of CORALSEA software. The models generated were based on hybrid optimal descriptors including Graphs and SMILES. All the models were observed to be robust via internal and external validation parameters. Model M4 of split 3 had the highest average correlation coefficient (r2 = 0.7209), average \({R}_{m}^{2}\) value (0.6256) and lowest error (MAE = 0.465). Structural features characterised as promoters and hinderers were extracted from the M4 model and their mechanistic interpretation was employed to improve the pIC50 of least potent compound SM33 from the dataset. Removal of hinderers and addition of promoters resulted in a series of compounds, DDSL6a-DDSL6l among which DDSL6j was the most active compound with a pIC50 of 8.07. All the compounds were subsequently subjected to molecular docking against PDB: 4ASD, with DDSL6j having better LF rank score (− 14.032) and LF dG score (− 12.318). MD simulation of DDSL6j, sorafenib and SM33 carried out at 200 ns revealed sustained interactions of these compounds with all the key amino acid residues of the protein. Improved potency of compound DDSL6j was attributed to presence of additional stable H-Bond interaction. MM/GBSA calculation also indicated stable complex with binding free energy of 0.28 kcal/mol. All the designed molecules were also found to be possess good ADMET profile. These findings indicate that the quinazoline nucleus can be investigated and refined to create novel VEGFR-2 inhibitors as potential anti-angiogenic drugs in future.
