Despite advances in rapid molecular techniques for tuberculosis (TB) diagnostics, there is an unmet need for a point-of-care, nonsputum-based test. Previously, through a T7 phage antigen display platform and immunoscreening, we identified that the serum IgGs of active TB patients differentially bind to several antigen-clones and that this immunoreactivity discriminates TB from other respiratory diseases. One of these high-performance clones has some homology to the transketolase of Mycobacterium tuberculosis (M.tb TKT). In this study, we developed a direct enzyme-linked immunosorbent assay (ELISA) detecting IgG against the TKT antigen-clone (TKTμ). Through sequence alignment and in silico analysis, we designed two more peptides with potential antigenicity that correspond to M.tb-specific transketolase (M.tb TKT1 and M.tb TKT3) epitopes. After the development and standardization of a direct peptide ELISA for three peptides, we tested 292 subjects, including TB (n = 101), latent tuberculosis infection (LTBI) (n = 49), healthy controls (n = 66), and sarcoidosis (n = 76). We randomly assigned 60% of the subjects to a training set to create optimal models to distinguish positive TB samples, and the remaining 40% were used to validate the diagnostic power of the IgG-based assays that were developed in the training set. Antibodies against M.tb TKT3 yielded the highest sensitivity (0.845), and these were followed by TKTμ (0.817) and M.tb TKT1 (0.732). The specificities obtained by TKTμ, M.tb TKT3, and M.tb TKT1 on the test sets were 1, 0.95, and 0.875, respectively. The model using TKTμ obtained a perfect positive predictive value (PPV) of 1, and this was followed by M.tb TKT3 (0.968) and M.tb TKT1 (0.912). These results show that IgG antibodies against transketolase can discriminate active TB against LTBI, sarcoidosis, and controls.