Predictive QSAR modeling of HIV reverse transcriptase inhibitor TIBO derivatives

摘要: Abstract Comparative quantitative structure–activity relationship (QSAR) studies have been carried out on tetrahydroimidazo[4,5,1- jk ][1,4]benzodiazepine (TIBO) derivatives as reverse transcriptase inhibitors ( n  = 70) using topological, structural, physicochemical, electronic and spatial descriptors. The data set was divided into training test sets a cluster-based method. Linear models were developed multiple regression (with stepwise regression, factor analysis genetic function approximation (GFA) variable selection tools) partial least squares (PLS) combination of (FA–PLS). Genetic (spline) artificial neural networks (ANN) used for the development non-linear models. Using topological structural descriptors, best equation obtained from GFA based internal validation Q 2  = 0.737), but model with external characteristics FA–PLS R pred  = 0.707). When descriptors used, (0.740) value whereas PLS provided (0.784) value. all in combination, (0.760) (0.800) ANN (spline), respectively. majority satisfied criteria recommended by Golbraikh Tropsha (2002) modified r m ) values suggested Roy (2008). In order to further validate selected models, an 10 TIBO derivatives, which fall within applicability domain are not shared compounds present set, taken different source, inhibitory activity these predicted. Acceptable squared correlation coefficients between observed predicted suggesting true predictive potential