Chloramphenicol Borrman classification br Ming classificatio
TWEAK 0·0072 WHO classification
of the top 19 Chloramphenicol showed the optimal AUC accuracy number when no significant improvement was found by adding one more protein to the linear combination (Fig. 3A and B). Therefore, the generated diag-nostic signature was found to be:
The sensitivity and specificity of the combined model for distinguishing GC patients from controls were 93% and 100%, respec-tively, with an AUC of 0.99 (95% CI: 0·98–1). The diagnostic perfor-mance of the combined signature is clearly greater than that of each of the 19 proteins and also the clinical used biomarkers CA19–9 and CA72–4 (Table 3). Querying the STRING database confirmed the ab-sence of experimental evidence of protein-protein interactions among those selected 19 proteins. Only MMP1, and MMP10 and other MMP
family members may form a complex which activates MMP9; MSLN and CEACAM5 are both attached to cell membrane via glycosylphosphatidylinositol (GPI) anchors (Fig. 3C).
3.8. Proteins significantly altered in sera from GC patients at TNM I-II early stage
Cancer patients at early stage are always difficult to diagnose but early detection is important for successful therapy. Twenty-eight GC pa-tients were diagnosed at the earlier TNM I-II stage in the present cohort. Volcano plot in Fig. 4A illustrates the significantly altered proteins be-tween patients at early stage and controls by univariate analysis. GCNT1 was shown as the most significantly differential protein, and its optimal diagnostic sensitivity, specificity, and AUC of GCNT1 in pa-tients at TNM I-II stage determined by ROC curve were 75%, 86% and 0·82, respectively (Supplementary Fig. 10A and B). PCA plots for both the distribution of tissue and serum samples according to TNM stages as well as volcano plots for protein alterations in different group com-parisons in both tissue and serum are demonstrated in Supplementary Fig. 11. With the 19-serum protein signature identified above for the whole cohort, the diagnostic performance for differentiating patients at TNM I-II stage from controls was better than that of each individual protein with an AUC of 0·99 and sensitivity of 89% and specificity of 100% (Fig. 4B), whereas the best score for a single protein was for MMP1 with a sensitivity of 68%, specificity of 78% and AUC of 0·75, and the AUCs for clinically measured biomarkers CEA, CA19–9, and CA72–4 were 0.58, 0.48, and 0.61, respectively.
3.9. Proteins significantly altered in sera from GC patients with high micro-satellite instability
One of the leading causes of GC is a defect in DNA mismatch repair, resulting in microsatellite instability. Microsatellite-unstable tumours are hyper-mutated intestinal subtype tumours and have recently been proposed as one of the most robust subgroups in molecular characteri-zation of GC, which occurring in at least 20% of all GC patients and hav-ing a better overall prognosis and lower frequency of recurrence [26,27]. The molecular diagnosis of MSI GC is important before treatment, as it triggers different response to chemotherapy, and may require specific surgical treatment such as tailored lymphadenectomy, and it stratifies patients for targeted therapies . In the present cohort, only 35 GC patients were found with high MSI. As shown in Fig. 4C, the volcano plot displays the significantly changed proteins in serum of GC patients with MSI-H status when comparing to the controls by univariate analy-sis. GCNT1 was also the most significant altered protein in MSI-H versus controls, and the optimal diagnostic sensitivity, specificity, and AUC of GCNT1 for patients with high MSI status were 71%, 86% and 0·82, re-spectively (Supplementary Fig. 10C and D). Supplementary Fig. 12 illus-trates the PCA plots for sample distribution according to MSI status as well as volcano plots for protein alterations in different group compari-sons in both tissue and serum. The diagnostic performance of the 19-serum protein signature for differentiating patients with high MSI from controls was significantly better than that of each individual pro-tein with an AUC of 0·99 and sensitivity of 91% and specificity of 98%