Abstract: The optimization of geometries and electronic structures of 20 TIBO (Tetrahy-droimidazobenzodiazepinone) HIV-1 inhibitors had been performed by MNDO and ab initio quantum chemical methods. Asignificant QSARwas obtained as below:P_c= -10. 425+4. 216×10^-2V-19. 935Q_N12-4. 211Q_3x. The BP artificial neural network method can be used to predict the activities more accurately. The results show that (1) the activities of TIBO derivations will increase due to the larger volume and less polarity of molecules; (2) the hydrogen atom connected to nitrogen atom in ring Cmay be the active atom taking positive charge; (3) the R' atom(Sor O) is the electron donor to acceptor. The frontier molecular orbital component of Sis much bigger than that of O, and Sis more puweiful election donor than O, which indicate why the sulfur has a better activity than oxygen; (4) the electron acceptor substituents in ring Acan enhance the activity of TIBOderivations. Since smaller distance between positive and negative charge center leads to small polarity, 3-substituted derivations show better activities than 2-substituted derivations.

Key words: TIBO derivative, RT inhibitors, Ab initio, BP artificial neural network