Abstract: Objective 　To explore the screening of lung cancer with Bayes model of GSH-PX, MDA and oncogene proteins. Methods 　plasma lipid peroxidation from 48 cases of lung cancer (LC), 34 cases of non lung cancer of respiratory diseases (cont rol group) were measured with glutathione peroxidase ( GSH-PX) and malonydialdehyde (MDA) . Western Dot Blotting was used to explore the expression of ras, p53 and heat st ress protein 70 (hsp70) in LC and cont rols. Results 　The GSH-PX activity of LC was lower than that of cont rols, P < 0. 01. The level of MDA increased more significantly in LC group than in control group, P < 0. 05. The level of p21, p53, and HSP70 in LC group were higher than those in control group, P < 0. 01. The sensitivity, specificity and the predicative rate of Bayes model were 83. 33 %, 74. 19 % and 93. 33 % respectively. Co2detection of the model and conventional cytology produced a sensitivity of 91. 67 %. Conclusion 　Bayes model of GSH-PX, MDA, p21, p53 and HSP70 might be used as the screening of lung cancer and added to the diagnostic value of conventional cytology.