Correlation between mRNAsi and clinical characteristics in GAC
As shown, when compared with normal samples, mRNAsi of tumor samples was significantly higher (Fig. 1a) and the result of multivariate logistic regression analysis, adjusted by age, gender, approved it with P<0.001. Further, in terms of clinical features, 323 GAC patients were classified by age, gender, TNM stage, tumor stage, histopathological types, respectively; for which, mRNAsi was not associated with age (P = 0.133), gender (P = 0.780), T (P = 0.140), N (P = 0.579), M (P = 0.598) stage, but only had a declining trend with the improved tumor stage (Fig. 1b-g), and all the multivariate logistic regression results showed that the above analysis results were not statistically significant with P>0.05. However, in terms of histological subtype, mRNAsi in the intestinal-type was significantly higher than that in diffuse-type (Fig. 1h) and multivariate logistic regression analysis approved it with P=0.04. Finally, survival analysis indicated that the overall survival of patients with IGC and DGC made no sense (Fig. 1i, j) and multivariate Cox proportional hazards models also supportted this result with P>0.05 of both.
Screening of DEGs and stemness-related modules and key genes in IGC and DGC
The above results indicated that there may be specific DEGs that control the stemness of IGC and DGC, respectively. Therefore, we further filtered the significant modules and key genes with stemness properties. First, we compared the DEGs between IGC or DGC and normal samples, respectively. In IGC, 7596 DEGs were discerned, among which 5989 were over-expressed and 1607 were down-regulated (Fig. 2a and Additional file 1: Table S1); in DGC, 5424 DEGs were discerned, among which 4647 were up-regulated and 777 were down-regulated (Fig. 3a and Additional file 1: Table S2).
Furthermore, WGCNA was used to construct a co-expression network of DEGs with the soft threshold β= 5 (scale-free R2 = 0.90) or β= 8 (scale-free R2 = 0.90) to ensure a scale-free network, and 27 and 10 gene modules with similar expression profiles in IGC and DGC were obtained for subsequent analysis (Fig. 2b, 2c; Fig. 3b, 3c). Then, we used MS as the overall gene expression level of the corresponding module to calculate the correlations between the modules and mRNAsi in IGC and DGC. Here, we noted that the blue module of intestinal-type with a correlation close to 0.8 and the brown module of diffuse-type with a correlation close to 1.0, both showed a positive correlation between stemness properties and gene expression. In addition, the dark-gray and royal-blue modules of IGC, and black and red modules of DGC exhibited relatively high negative correlations with mRNAsi. Thus, we chose the blue module of IGC and brown module of DGC as the modules of greatest interest. The thresholds of screening key genes in the significant modules were defined as cor.GS> 0.5, cor. MM> 0.8 and cor. GS> 0.8, cor. MM> 0.8 respectively. Finally, 16 stemness-related genes including ORC6, BUB1, NCAPH, ORC1, WDHD1, RACGAP1, CKAP2L, RBL1, KIF18A, TTK, TPX2, MAD2L1, NCAPG, RAD54L, EXO1, PLK4 have obtained for intestinal-type, as shown in Fig. 2d-f; 43 key genes for the diffuse-type, including DLGAP5, DTL, NCAPG2, KIF11, NCAPH, EZH2, ORC6, MTHFD2, BUB1, RAD54L, XRCC2, BUB1B, HMMR, KIF18A, KIF2C, NCAPG,EME1,PLK4,GINS1,PARPBP,SPAG5,ZWILCH, RAD51AP1, RACGAP1, CCDC138, TPX2, SPC25, MAD2L1, CCNB2,DKC1,KNSTRN,ZWINT,RAD51, NUSAP1, CHAF1A, SGO1, FEN1, CCNB1, RRM2, CDCA8, MND1, CCT6A, DBF4, as shown in Fig. 3d-f. Then we plotted their expression tendency in normal and tumor samples and discovered that the candidate genes were overexpressed in IGC and DGC, respectively (Fig. 4a, 4b). However, taking the intersection of key genes involved in IGC and DGC, BUB1, KIF18A, MAD2L1, NCAPG, RAD54L and PLK4 were left, which were more highly expressed in IGC.
Validation of stemness-related key genes in the Oncomine database
Next, the mRNAsi-related key genes were verified in the Oncomine database. As shown in Fig. 5, 16 genes of IGC were highly expressed. And except for CHAFIA, 42 key genes showed higher expression in DGC. Fig. 6 showed 7 representative genes in DGC compared with normal tissues, and the remaining were shown in Supplementary Figure S1.
The cellular functions and pathway analysis of stemness-related key genes
In IGC, the GO analysis results indicated that key genes participated in the biological process of nuclear division. The main cellular component manifested enrichment mainly at the spindle. The main molecular function enriched these genes in DNA replication origin binding, and the KEGG analysis demonstrated that the major pathway was cell cycle (Fig. 7a, 7c). In DGC, GO analysis results indicated that key genes participated in the biological process of nuclear division. The major cellular component suggested enrichment mainly in the chromosomal region. And the main molecular function was catalytic activity. KEGG analysis showed that the main pathway was also the cell cycle (Fig. 7b, 7d).
Correlation between stemness-related key genes at transcription and protein levels
We explored the interactive relation of the above key genes using Pearson correlation and found that the key genes in the blue module of IGC or the brown module of DGC had relatively strong correlation, with the minimum correlation coefficient of 0.41 and 0.58, and the correlation among genes was shown in Fig. 8a and Fig. 9, respectively. Next, we built the PPI network using the STRING online tool. As shown in Fig. 8b and Fig. 10a, the key genes of the two types of GC formed a close interaction relationship respectively. In IGC, the most crucial key protein was BUB1 (15 edges) (Figure 8c). The most important key proteins in DGC were CCNB1 (37 edges) and RAD51AP1 (37 edges) (Figure 10b) and BUB1 also had high connectivity.