Quantitative structure-activity relationship (QSAR) analysis of 28 quinoline-arylamidine (CQArA) hybrids against two leukemia cells, K562 and Raji, was performed. Multiple linear regression (MLR) models were obtained by genetic algorithm. The best models involved the following descriptors: radial distribution function (RDF) descriptors, GETAWAY (GEometry, Topology, and Atom-Weights AssemblY) descriptor, bond information content index, and dipole moment. The best MLR models for K562 and Raji cells demonstrated satisfactory stability in internal and external validation. Since the QSAR model for Raji cells has better predictive ability, two new highly potent CQArA analogues were proposed based on it. The QSAR models revealed important physicochemical and structural requirements for the antitumor activity: enhanced 3D molecular distribution of mass calculated at radius 11 Å from the center of molecule, a higher number of terminal electronegative atoms, extension of the molecules' central linker between quinoline and arylamidine, higher ratio of single bonds and total number of atoms, and symmetric charge distribution. Molecular docking study was applied to ensure the anticancer activity affinity to the binding site of the tyrosine-protein kinase (c-SRC). It was confirmed that the most active compound binds on the pocket between the small and large lobes of c-SRC, mostly throughout the hydrogen bonds and van der Waals interactions.