摘要:
We have applied Hierarchical QSAR Technology (HiT QSAR) approach employing simplex representation of molecular structure to the analysis of in vitro inhibitory effects of dual combinations of picornavirus replication inhibitors against poliovirus 1 (Mahoney). The possible dual combinations of arildone, disoxaril, enviroxime, guanidine, HBB, PTU-23, ribavirin and S-7 were investigated. Antiviral activity was expressed in lgIC50 (μM). Both molecules in the binary combinations were represented their simplexes. These representation allowed analyzing the synergy, antagonism or additivity in the combination's interaction with the biological target. We dubbed this approach as “double 2D QSAR”. Predictive QSAR models were obtained using PLS approach and validated using 8- fold external cross-validation (R2 ext = 0.67-0.93). Adequate models (R2 ext = 0.53-0.97) were obtained in the same way for IC30, IC40, IC60 and IC70 inhibitory concentrations. The usage of predicted values of IC30, IC40, IC60 and IC70 in the framework of the feature net approach allowed to increase the quality of our QSAR model (R2 ext = 0.71-0.94). Resulting QSAR models were analyzed, structural fragments and parts of combination promoting the antiviral activity were determined, e.g. oxypropylisoxasole. The developed model was then used to predict novel potent combinations of drugs. Oxoglaucine and pleconaril (which was very similar to disoxaril) were predicted as novel potent components of an inhibitory combination. HiT QSAR proved itself as an adequate tool for QSAR analysis of combinations and, although the method used is suitable only to binary combinations, it can be easily extended for more complex tasks.