Improved Genetic Algorithm (IGA), which has a higher efficiency than the conventional ones, has been used for FLARM (Flexible Atom Receptor Model) algorithm. All the atoms in a FLARM receptor model are spacially moveable in the process of genetic evolving. This can avoid the bias of the initial coordinate setting. Void atom is specially favored (a chance of 50%) in the atom selection to obtain open receptors model that allow large gaps. An open model may provide a better simulation of the real receptor and may rationally correspond to the pharmacphore model. Two data sets (one is the binding affinity to corticosteroid binding globulin of 21 steroids, and another is the anti-HIV-1 activities of 38 1,1,3-trioxo-2H,4H-thieno[3,4-e] [1,2,4] thiadiazine derivatives) were investigated by using FLARM. The resulted receptor models showed high prediction ability (q~2 are 0.782 and 0.72 for the predicting sets of the two systems, respectively) that can be used to predict the activity of new drugs. Moreover, The explicit atom models can help to understand the mechanism of the interaction between ligands and receptors and even to preliminarily predict the active sites of the real receptors.

Key words: STEROIDS, THIOPHENE P, THIADIAZINE P, BIOLOGICAL ACTIVITY