Constructing A Risk-scoring Model
By univariate Cox analysis, we identified 115 anoikis-related genes associated with prognosis (Fig. 1A). Seven anoikis-related genes were screened by LASSO regression analysis (MET, DYNLL2, CDK1, TNFSF10, PIP5K1C, MSLN, GKN1) (Fig. 1B-C). We also analyzed the PPI network of these 7 genes by string (Fig. 1D). We constructed a risk score model based on seven anoikis-related genes and divided the patients into high-risk and low-risk groups by the median. Kaplan-Meier survival analysis displayed that the overall survival time of patients in the high-risk group was significantly lower than that in the low-risk group (Fig. 2A). The survival and risk score of each osteosarcoma patient are shown in Fig. 2B-C. The gene heat plot shows the expression level of seven genes in each patient (Fig. 2D). Finally, we estimated the accuracy of the risk score model by the time ROC curve, and the results displayed that the AUC (area under the curve) values of patients at 1, 2, and 3 years were 0.754, 0.746, and 0.774, respectively (Fig. 2E). In the GEO validation group, we obtained similar results: the overall survival time of osteosarcoma patients in the low-risk group was longer than that of osteosarcoma patients in the high-risk group (Fig. 3A-D). The AUC values of 1, 2, and 3 years ROC curves of GEO patients were 0.705, 0.684, and 0.781, respectively (Fig. 3E).
Proteomics
We found significant P values for EGFR and MMP2 proteins by univariate and multifactorial cox analysis (along with risk scores) (Fig. 4A-B). Kaplan-Meier survival analysis shows survival grouped by median EGFR and MMP2 protein expression (Fig. 4C-D). Interestingly, both EGFR and MMP2 proteins are anoikis-related genes expressed proteins. This is further evidence that anoikis-related genes have a very important impact on pancreatic adenocarcinoma development.
Clustering And Analysis Of Somatic Mutations
We performed cluster analysis based on seven prognostic genes and found that they could be clustered into three categories (Fig. 5A). Category 3 patients had the best survival and the lowest risk scores. Category 2 patients had the worst survival, but the risk scores of category 2 patients and category 1 patients statistically showed no difference (Fig. 5B-C). We analyzed the somatic mutations in the three categories of patients in order to further investigate the reasons for the differences in survival. The results showed that patients in category 3 with the best survival had significantly fewer KRAS, P53 and CDKN2A mutations (Fig. 5D-F). We grouped patients according to the presence of KRAS, P53 and CDKN2A mutations, and Kaplan-Meier survival analysis showed that patients with KRAS, P53 and CDKN2A mutations had poorer survival (Fig. 5G-I). We show the percentage of KRAS, P53 and CDKN2A mutations in the three categories of patients in a bar chart. The results show that patients in category 3 with the best survival have the lowest percentage of all three mutations, while patients in category 2 with the worst survival have the highest percentage of CDKN2A mutations (Fig. 5J-L). Therefore, speculate that the worst survival in class 2 patients may be associated with CDKN2A mutations, although this may require further confirmation. We drew Sankey plots based on the three clusters of patients, CDKN2A mutation and high and low-risk groups, which were used to clearly show the correlation between the three (Fig. 5M).
Immune Infiltration In The Low-risk And High-risk Groups
In patients with TCGA pancreatic adenocarcinoma, CIBERSORT analysis showed that compared with the high-risk group, the low-risk group had a higher proportion of T-cells CD8 and T cells gamma delta, while a lower proportion of Macrophages M0 and Macrophages M1 (Fig. 6A). ssGSEA analysis showed that compared with the high-risk group, the low-risk group had a higher content of Activated B cell, Activated CD8 T cell, Activated dendritic cell, Central memory CD4 T cell, Eosinophil, Macrophage, Mast cell, MDSC, Monocyte, Plasmacytoid dendritic cell, T follicular helper cell, Type 1 T helper cell and Type 2 T helper cell, while a lower proportion of Activated CD4 T cell (Fig. 6B). Finally, compared with the low-risk group, the high-risk group had lower expression of the immune checkpoint, suggesting that the low-risk group was more suitable for immune checkpoint inhibitor therapy (Fig. 6C-D). In patients with GSE57495, CIBERSORT analysis showed that compared with the high-risk group, the low-risk group had a higher proportion of T-cells CD8, T cells gamma delta, while a lower proportion of Macrophages M0, Dendritic cells resting, Eosinophils and Neutrophils (Figure S1A). ssGSEA analysis showed that compared with the high-risk group, the low-risk group had a higher content of Activated B cell, Activated CD4 T cell, Effector memeory CD8 T cell, Mast cell, MDSC, T follicular helper cell and Type 1 T helper cell, while a lower proportion of Activated CD4 T cell, Activated dendritic cell, Gamma delta T cell, Neutrophil, Type 17 T helper cell and Type 2 T helper cell (Figure S1B). Unlike TCGA, we did not find high expression of immune checkpoints in the low-risk group. Instead, the high-risk group overexpressed CD80 (Figure S1C-D).
Enrichment Analysis
Figure 7A shows the volcano map of differential genes in the high-risk and low-risk groups. GSEA enrichment analysis displayed that the differential genes were related to
intracellular transport, insulin secretion regulation, neurotransmitter secretion, signal release, transport vesicle and vesicle mediated transport in synapse (Fig. 7B). GO enrichment analysis showed the higher involvement of the differential genes in the regulation of ion transmembrane transport, signal release, presynapse and channel activity (Fig. 7C). KEGG enrichment analysis displayed the relation between the differential memory to the neuroactive ligand − receptor interaction, cAMP signaling pathway, Pancreatic secretion and other pathways (Fig. 7D).
Validation In Single Cell Sequencing
We finally obtained 28 cell clusters and annotated the cell type of each cluster (Fig. 8A). MET and TNFSF10 is mainly expressed in epithelial cells and endothelial cells. MSLN is mainly expressed in epithelial cells. CDK1 is mainly expressed in mesenchymal stem cells (MSC) (Fig. 8B).
Drugs With Potential Efficacy In Pancreatic Adenocarcinoma
Drug sensitivity analysis showed that gemcitabine showed significant sensitivity in the low-risk and high-risk groups. Patients in the high-risk group were more sensitive to gemcitabine, with a correlation coefficient of -0.36 between risk score and IC50 of gemcitabine (Fig. 9A-B).
Real-time Quantitative Pcr
Expression levels of the anoikis-related genes in normal and pancreatic adenocarcinoma cells were measured. As shown in Fig. 10, compared with normal esophageal samples, the expression levels of MET, DYNLL2, CDK1, TNFSF10, PIP5K1C, MSLN and GKN1 were significantly increased in pancreatic adenocarcinoma cells (p < 0.05).