• 2019-07
  • 2019-08
  • 2019-09
  • 2019-10
  • 2019-11
  • 2020-03
  • 2020-07
  • 2020-08
  • 2021-03
  • br Results br Genome wide


    3. Results
    3.1. Genome-wide discovery of a novel gene Ac-DEVD-CHO signature to detect lymph node metastasis in early stage gastric cancer
    subsequently derived a LN risk scoring formula. To demonstrate the robustness of this panel as a diagnostic marker, and its applicability to patients with T2 stage GC, we first evaluated its performance in the training set of T1 patients. This was followed by in silico validation in an expanded TCGA dataset involving 96 T1/T2 patients (LNP 49, LNN 47), and another independent set of 188 T2 patients (LNP 157, LNN 31) from the ACRG cohort.
    In the clinical validation phase, two large, independent patient cohorts were analyzed to validate the 15-gene signature identified during the discovery phase. Using qRT-PCR data derived from 101 T1/T2 patient (LNP 24, LNN 77) specimens in the clinical cohort-1 as the testing set, we conducted a multivariate logistic regression analysis for qRT-PCR, from which a LN risk scoring formula was derived. The diagnostic performance of the 15-gene signature subsequently evaluated using an independent qRT-PCR dataset from 147 (LNP 26, LNN 121) T1 specimens from the clinical validation cohort-2. To demon-strate the clinical significance of our data, we benchmarked our gene signature against the conventional tumor markers, CEA and CA19–9. Computed tomography (CT) was performed before surgery in all pa-tients belonging to the clinical cohort-2, and the imaging results were evaluated by board certified radiologists. When the size on the short axis of the regional LN was N10 mm, clinical LN status was deemed to be positive.
    2.4. RNA isolation and quantitative reverse-transcription PCR
    Total RNA extraction from tissue specimens was performed using miRNeasy RNA isolation kits (Qiagen, Hilden, Germany). Synthesis of complementary DNA (cDNA) was conducted on 1 μg of total RNA using the High Capacity cDNA Reverse Transcription Kit (Invitrogen,  To identify a panel of genes that can help diagnose patients with lymph node metastasis, we first analyzed RNA-seq expression profiling data from 18 patients with early stage T1 cancers, which were either LN metastasis positive or negative. Among a total of 20,531 genes, 84 genes were differentially expressed between 5 lymph node positive (LNP) and 13 negative (LNN) patients (P b .01 [Wilcoxon signed-rank test], log2 fold change N1.5; Fig. 1a and S2). To identify a robust candidate gene sig-nature, we further narrowed down the gene list to 15 by filtering out lowly expressed genes (average expression level b 3 log2-transformed TPM). Using multivariate logistic regression analysis, we found the 15 candidate genes were able to successfully distinguish LNP from LNN GC patients in the training set (AUC = 1.000, 95% CI 1.000–1.000; Fig. 1b).
    In view of the availability of multiple public datasets consisting of T2 GC patients, we next investigated whether our T1 lymph node metasta-sis GC gene signature could also identify LN status in these additional patient cohorts. Intriguingly, our genes were able to distinguish LNP from LNN patients in an expanded set of 96 T1 and T2 patients in the TCGA cohort (AUC = 0.839, 95% CI 0.757–0.921; Fig. 1c), as well as in the ACRG cohort of 188 T2 patients (AUC = 0.829, 95% CI 0.752–0.906; Fig. 1d). These data highlight the diagnostic potential of our novel 15-genes in identifying LN metastasis in early-stage gastric cancer patients.
    3.2. Validation and establishment of the 15-gene signature for detecting Lymph node status in gastric cancer patients
    Next, we assessed the diagnostic accuracy of the 15 gene-panel by qRT-PCR in 24 LNP and 77 LNN tissue specimens in the clinical cohort-1 (n = 101). While individual genes had limited predictive power (AUCs varying between 0.506 and 0.605), the combination of
    Of interest, our 15-gene signature was significantly superior com-pared to the conventional tumor markers, CEA (P = .033 [DeLong]) and CA19–9 (P = .044 [Delong]; Fig. 2b) in identifying LNP patients. To further validate the diagnostic efficiency of this 15-gene signature, we next examined its performance in an independent validation cohort comprising of 26 LNP and 121 LNN T1 GC patients using the multivariate logistic regression analysis. In line with results from our testing cohort, the 15-gene signature was once again able to robustly distinguish LNP from LNN early stage GC patients (AUC = 0.742, 95% CI, 0.631–0.852; OR = 6.563, 95% CI, 2.585–16.66; Fig. 2b and Table 2).