Background The extended thymectomy for myasthenia gravis (MG) is currently available, but in 20%-40% of the patients the results were not satisfactory. There are no ideal indicators forecasting surgical results before operation. The surface enhanced laser desorption ionization-time of flight-mass spectroscopy (SELDI-TOF-MS) is a currently new technique for detection of protein profiles, and some progresses have been made in cancer diagnosis and efficacy evaluation, but there is no report on efficacy forecasting of MG surgery. This study aimed to establish an efficacy prognosis model for forecasting the efficacy of surgery for MG by analysis of serum protein profiles of MG patients before surgery. Methods Fifty-six MG patients 6 months after extended thymectomy were enrolled in the study. They were classified into effective or non-effective groups according to symptoms and medication. Their pre-operative blood samples were analyzed for protein profiles by the SELDI-TOF MS technique, and protein peaks were identified for establishment of the efficacy prognosis model of MG surgery. Additional 100 MG patients were subjected to model validation and their pre-operation protein profiles reviewed for post-operative results. The results were compared with those of the post-operative follow-up so as to validate the prognosis model. Results For the model establishment, symptoms were improved in 33 patients and not improved in 18 patients, with an effective rate of 64.7%. Five (8.9%) patients were lost to follow-up. Within the molecular weight range of 1 000 to 20 000, 3 specific protein peaks were found to be significantly different between the effective and non-effective groups, ie M4110-76, M3394-58, and M1258-55. Using the efficacy prognosis model constructed with these data, the accuracy rate of classification was 87.9% for the effective group, and 83.3% for the non-effective group, with a total accuracy rate of 86.3%. For the model evaluation, 2 (8.9%) patients were lost to follow-up, 62 patients were effective and 36 were non-effective. By comparing with the real results of follow-up with 65 effective and 33 non-effective patients with an effective rate of 66.3%, the accuracy rate of prediction by the prognosis model was 86.2% for the effective group, and was 81.8% for the non-effective group with a total accuracy rate of 84.5%. Conclusions By protein profiles analysis of pre-operative blood samples taken from MG patients with the SELDI-TOFMS technique, protein peaks correlated with surgery efficacy in MG patients can be found for primary forecasting short-term efficacy of surgery for MG patients.