| Background: gastrointestinal cancer has become an important threat to human life and health due to its increasing incidence and poor prognosis.Colorectal cancer and stomach cancer account for 17% of new cases of tumor and cause 1.69 million deaths worldwide each year.Although the diagnosis and treatment technology of gastric cancer and colorectal cancer has been continuously improved.Due to the high heterogeneity and complicated mechanism of gastrointestinal caner,there are still a large number of patients with gastric cancer and colorectal cancer who have experienced recurrence and metastasis after receiving standardized radical treatment,especially for advanced gastric cancer and colorectal cancer.With the development of bioinformatics technology,it has become possible to obtain multidimensional omics information of tumor and peripheral tissue in a high throughput way.Molecular typing the gastrointestinal cancer based on its heterogeneity information obtained from multi-omics study could be an essential approach for the accurate treatment.Aim: Develop a bioinformatics algorithm suitable for the screening and identification of gastrointestinal tumors’ prognostic biomarkers.Identify a optimal set of several prognostic genes and establish the corresponding scoring system which can group the prognosis of patients with gastric or colorectal cancer.Verify the reliability and independence of the prognostic scoring system in large-scale clinical paraffin specimens by novel assay based on nucleic acid hybridization and explore its guidance value for clinical diagnosis and treatment decision5 Methods: In the part of colorectal cancer,we used multiple independent public databases to identify the optimal gene set related to colorectal cancer prognosis through a multi-step bioinformatics algorithm.The efficacy of our 15-gene colorectal cancer prognosis scoring system to Oncotype DX Colon Recurrence Score in prognosis prediction of colorectal cancer patients was compared in several independent dataset.Furthermore,we applied a novel gene expression level detection technique based on nucleic acid molecular hybridization to obtain the expression levels of prognostic genes in 203 paraffin specimens after colorectal cancer radical surgery in one single center.Cox regression and Kaplan-Meier analysis were employed to verify the efficacy and independence among the clinical samples.In the part of gastric cancer,we employed the similar approaches described above to validate our 53-gene gastric cancer prognosis scoring system.540 paraffin specimens after gastric cancer radical surgery were obtained from 3 clinical centers for clinical validation.We found our 53 gene gastric cancer prognosis scoring system has the guidance value for chemotherapy.Results: In the part of colorectal cancer,we identified 738 genes with consistently high or low expression between tumor and normal tissues from 6 independent colorectal cancer datasets.Among them,78 genes(P<0.05)that had significant influence on the prognosis of colorectal cancer were screened out.Through a series of bioinformatics algorithms,15 optimal prognostic genes were selected and scoring system were established.The predictive efficacy of our scoring system was validated among one training set(373 samples)and two independent validation sets(125 and 562 samples)and P value were 0.0001,0.0033 and 0.00058,respectively.The predictive efficacy of the 15-gene prognostic scoring systems was independent with the existing clinicopathological and molecular subtypes(P<0.0001).Compared with existing colorectal cancer prognostic scoring systems,the hazardous ratio of our scoring system between "poor" and "good" groups was 2.32-fold and 1.58-fold,in 2 validation datasets separately.Among the 203 samples from single center clinical validation,the OS of three prognostic groups differed significantly(P<0.0001).In the part of gastric cancer,we compared the efficacy of 53 gene prognostic scoring system for gastric cancer with published gene markers.The hazardous ratio of our scoring system between "poor" and "good" groups was 4.30-fold,1.48-fold and 1.40-fold higher than other 3 published gene biomarkers.The multivariate Cox regression analysis including molecular subtypes and clinicopathological information from public databases and clinical samples revealed that our prognostic scoring system was independent of these factors(P <0.0001 for all).We validated the 53 gene prognostic scoring system among 540 samples from three clinical centers.The P value in in the three independent validation sets and all samples were 0.037,0.043,0.21 and 0.0018,separately.Combined with chemotherapy regimens,we found that patients receiving FOLFOX regimen had a better 5-year survival rate than other regimens,FOLFOX group: 82.0%,other regimens: 61%(P=0.028).Conclusion: We developed a multi-step bioinformatic algorithms combining multi-step regression and validation models.It is capable of identifying the optimal prognostic gene set in gastrointestinal cancer and establishing a reliable prognostic scoring system based on this.In the part of colorectal cancer,we proved that our colorectal cancer prognosis scoring system was not only independent of the existing molecular subtype and clinicopathological factors of colorectal cancer,but also superior to the Oncotype DX Colon Recurrence Score in the field of prognosis prediction.In the part of gastric cancer,we confirmed that our 53-gene prognostic scoring system previously developed has a higher independent predictive efficacy compared with the published prognostic gene biomarkers of gastric cancer.we also demonstrated that our novel assay based on nucleic acid hybridization could enable a high throughput detection of gene expression levels in paraffin specimens.Furthermore,our colorectal and gastric cancer prognostic scoring systems have been demonstrated a prognostic ability in large-scale single-center or multi-center clinical validation.We also found that 53-gene gastric cancer prognostic scoring system has potential predictive value for the efficacy of FOLFOX. |
PDF Full Text Request