Abstract
We have discovered a new class of pyrido[b]bindole derivatives that show potent and broad spectrum anticancer activity with IC(50) values down to submicromolar levels. Structure-activity relationship data acquired with the compounds as antiproliferative agents against several cancer cell lines, i.e. human HCT116 colon cancer cell line, and HPAC, Mia-PaCa2 and Panc-1 pancreatic cancer cell lines, were subjected to two different QSAR modeling methods. A kernel-based partial least squares (KPLS) regression analysis with chemical 2D fingerprint descriptors, and a PHASE pharmacophore alignment with 3D-QSAR study. The KPLS method afforded successful predictive QSAR models for antiproliferative activity of the HCT116 colon cell line and on two of the pancreatic cancer cell lines HPAC and Mia-PaCa2, with the following statistics: R (2)s of 0.99, 0.99 and 0.98, for training set coefficients of determination, and external test set predictive r (2)s of 0.70, 0.58 and 0.70, respectively. The best 2D fingerprint descriptor for both the HCT116 and HPAC data out of the eight finger prints utilized was the atom triplet fingerprint; whereas the one that worked best for the Mia-PaCa2 data was the linear fingerprint descriptor. The PHASE pharmacophore based 3D-QSAR study afforded a four-point pharmacophore model comprising one hydrogen bond donor (D) and three ring (R) elements, which yielded a successful 3D-QSAR model only with the HCT116 cell line data with training set R (2) of 0.683, and an external test set predictive r (2) of 0.562. With the PHASE 3D-QSAR, the influence of electronic effects and hydrophobicity were visualized, and were in agreement with the observed SAR of substitutions, while the KPLS method the relative extent of contribution of each atom in a compound to the activity. These models will foster the lead optimization process for this potent series of anticancer pyrido [3,4-b]indole compounds.