Gut microbiota community shift with coronary artery disease (CAD) has been reported in several limited cohorts during the past several years. However, whether the enriched or decreased microbiota taxa with CAD can be reproducible deserves further investigation and validation. In this study, 78 human subjects were recruited. Of these, 19 were diagnosed without stenosis in coronary artery (control group, referred to herein as Ctrl), 14 with stenosis less than 50% (LT50), and 45 with stenosis greater than 50% (GT50). Fecal samples were collected and DNA was extracted to perform 16S rRNA gene sequencing. The operational taxonomic units (OTUs) were analyzed to identify taxa specific to different groups; next, multivariate logistic regression was employed to test whether the defined taxa could independently predict CAD risk. We found that Deltaproteobacteria, Fusobacterium, Bilophila, Actinomyces, and Clostridium XIX were enriched in Ctrl; Prevotellaceae, Parabacteriodes, and Butyricicoccus were enriched in LT50; and Roseburia and Butyricimonas were enriched in GT50. Further analysis revealed that increased populations of Deltaproteobacteria, Fusobacterium, Bilophila, and Desulfovibrionaceae were associated with a 0.26-fold, 0.21-fold, 0.18-fold, and 0.26-fold decreased risk of CAD, respectively (p < 0.05), and an increased Prevotellaceae population was associated with a 5.63-fold increased risk of CAD (p < 0.01). A combination of the 20 microbial taxa achieved an area under the receiver operating characteristic (ROC) curve of higher than 0.88 for all discriminations between LT50 vs Ctrl, GT50 vs Ctrl, LT50 + GT50 vs Ctrl, and GT50 vs Ctrl + LT50. However, the microbial taxa previously reported as enriched in CAD patients or healthy controls could not be observed in our cohort except for Bacteroides. In conclusion, CAD patients showed a different microbial taxa signature than the healthy controls. However, the non-reproducibility of the microbiota taxa enriched in CAD across different cohorts limits the use of this signature in early diagnosis and prevention. Only decreased Bacteroides abundance was found to be a reliable marker to indicate CAD progression.