The 3 dimension structure-activity relationships (3D-SAR) for N1-site substituted fluoroquinolones that used as antibiotics were performed with RS (rough sets) method.The process of RS is a concise nonlinear process, it has broad application foreground in the data mining of nonlinear life course.In this paper, initially the parameters of N1-site substituted quinolones were computed with the semi-empirical method: AM1 (Austin model) and MNDO3 (modified neglect diat. overlap), and the information table is composed of the parameters (condition attributes) and biol. activity (decision attributes).According to the anal. of rough sets method, the core and reduction of attributes set were obtained, then the decision rules were extractedThe results indicate that LUMO, A5, A3 and dipole, etc. have the great contribution to the activity of N1-site substituted fluoroquinolones.Integrating the related biol. research of fluoroquinolones resistance, the binary hydrogen bond mechanism of fluoroquinolones is brought forward.It implies that fluoroquinolones realize their function as electron donors as well as electron acceptors in its function process, and the process is finished by the formation of two hydrogen bonds (6-site F and 3-site OH).The mutation of bacteria makes it hard to form hydrogen bonds, then cause the fluoroquinolones resistance.The result also indicates that moderate dipole and ΔHf are helpful to fluoroquinolones to be absorbed by the human body and act on bacteria.As a nonlinear system, RS theory can extract the special relation in the database, it has nonlinear character compared with PCA (Principal Component Anal.), PLS (Partial Least Square), etc., and the result has unambiguous phys. meaning compared with ANN (Artificial Neuron Network), etc.The result is very succinct and pointed, it is instructive to the study of pharmacody function process, resistance mechanism of fluoroquinolones and development of new fluoroquinolones.