Time-dependent AI-Modeling of the anticancer efficacy of synthesized gallic acid analogues

Abstract

Background/AimMain objective of this study is mapping of the anticancer efficacy of synthesized gallic acid analogues using modeling and artificial intelligence (AI) over a large range of concentrations and exposure times to explore the underline mechanisms of drug action and draw careful inferences regarding drug response heterogeneity. MethodsTwo series of gallic acid derivatives i.e. esters and amides have been synthesized and characterized by FTIR, NMR and mass spectrometry. The compounds have been tested in vitro for their anticancer activity against wild type human ovarian cancer cell line A2780, prostate cancer cell line PC3 and normal human fibroblast cells 3T3. To completely characterize optimal anticancer activity, a comprehensive model using piecewise recursive Hill model is used to quantitatively assess the in vitro anticancer effect of the tested compounds as a function of concentration and exposure time for periods ranging from 24 to 72 h. A robust artificial intelligence approach i.e. the “Support Vector Machine (SVM) Learning Algorithm” is adopted to utilize the data obtained at different temporal values, to identify compounds that trail forecasting algorithm. ResultsAll the synthesized analogues were found biocompatible. Significantly low EC50 values indicated that tested compounds have potent anticancer activity against A2780 cell line in comparison to PC3 cells where only few compounds generated same impact at almost 200 times high dose. On the basis of EC50 values, compounds 7 h, 7 m, 9c, 9b, 7c, 7b and 7 g were identified as the most active anticancer agent against A2780. Three major patterns of drug response heterogeneity were observed for different compounds in the form of multiple Hill graphs and shallow slopes. The anticancer efficiency of the compounds was verified using Machine learning SVM regression learner algorithm. For compounds 7a, 7b, 7e, 7 g, 7o, 7 r, 9b, 9e-9 g higher accuracy was found in predicted and experimentally obtained end point potency in terms of % viability. ConclusionsPharmacodynamics modeling of anticancer potential of the synthesized compounds revealed that drug efficacy and response heterogeneity could be modulated by changing the exposure time to optimize therapeutic impact. Combining experimental results with AI based drug action forecasting, compounds 7b, 7 g, and 9b may be tested further as potent anticancer agent for in vivo studies. This approach may serve a useful tool for extrapolation of in vitro results for generating lead compounds in in vivo and preclinical studies.