The focus of this study was to detect novel sera biomarkers for smear-positive and smear-negative pulmonary tuberculosis and to establish respective diagnostic models using the surface-enhanced laser desorption ionization time-of-flight mass spectrometry (SELDI-TOF MS) technique. A total of 155 sera samples from smear-positive pulmonary tuberculosis (SPPTB) and smear-negative pulmonary tuberculosis (SNPTB) patients and non-tuberculosis (non-TB) controls were analyzed with SELDI-TOF MS. The study was divided into a preliminary training set and a blinded testing set. A classification tree of spectra derived from 31 SPPTB patients, 22 SNPTB patients, and 42 non-TB controls were used to develop an optimal classification tree that discriminated them respectively in the training set. Then, the validity of the classification tree was challenged with another independent blinded testing set, which included 20 SPPTB patients, 14 SNPTB patients, and 26 non-TB controls. SNPTB patients and non-TB controls also were analyzed alone using the same method. The optimal decision tree model with a panel of nine biomarkers with mass:charge ratios (m/z) of 4821.45, 3443.22, 9284.93, 4473.86, 4702.84, 3443.22, 5343.26, 3398.27, and 3193.61 determined in the training set could detect 93.55%, 95.46%, and 88.09% accuracy for classifying SPPTB patients, SNPTB patients, and non-TB controls specimens, respectively. Validation of an independent, blinded testing set gave an accuracy of 80.77% for controls, 75.00% for SPPTB, and 71.43% for SNPTB samples using the same classification tree. With the peaks displaying differences between SNPTB patients and non-TB controls, a simplified dendrogram (m/z 4821.45, 4792.74) demonstrated classification efficacy of 85.94% (sensitivity 86.36% and specificity 85.71%) for distinguishing SNPTB patients from non-TB controls. The independent blinded testing set containing 14 SNPTB patients and 26 non-TB controls gained an accuracy of 81.59% (sensitivity 78.57% and specificity 84.62%) for diagnosing SNPTB. Special proteins/peptides may change in SPPTB and SNPTB patients and those changes may be used to distinguish them with the proper discriminant analytical method and to pursue and identify some involved proteins underlying the biological process of tuberculosis.