Font Size: a A A

Screening For Cervical Cancer Potential Based On Differential Analysis Prognostic Biomarkers

Posted on:2024-08-17Degree:MasterType:Thesis
Country:ChinaCandidate:M ZhuFull Text:PDF
GTID:2544307073971369Subject:Applied statistics
Abstract/Summary:PDF Full Text Request
Cervical cancer occupies a place in the list of malignant tumors in the world,and patients with cervical cancer have a high incidence and mortality.Although in recent years,the awareness of timely HPV vaccination among school-age women has greatly helped to effectively prevent cervical cancer,China’s vast territory easily leads to factors such as uneven regional development,which directly leads to a serious health problem,that is,in the next few decades,cervical cancer will still exist and exist for a long time,and there are still potential problems in the prognosis of cervical cancer patients.Cervical cancer patients often have recurrent illness or cancer metastasis in the later stage,which is very unfavorable for the treatment space of patients in the later stage,re sulting in poor prognosis of patients.We need to look for biomarkers with potential prognostic value as much as possible,specifically considering the aspects of cervical cancer recurrence monitoring,prognosis evaluation and individualization that is conducive to accurate treatment of patients.According to relevant data,early diagnosis and treatment of cervical cancer patients are closely related to reducing its morbidity and mortality.In recent years,no suitable and specific biomarkers have been found in clinical diagnosis,treatment and prognosis detection of cervical cancer.Although some biomarkers with potential value have been tested in experiments,the clinical results of sensitivity or specificity tests are not very satisfactory,so we urgently need to explore biomarkers with prognostic value for diagnosis,treatment,recurrence monitoring and prognosis evaluation.In this paper,based on the method of using differences,according to the data provided by the existing public databases,through standardized processing,we build a model to find molecular biological markers related to the diagnosis,treatment and prognosis of cervical cancer,explore the differential genes and related signal pathways that may affect the prognosis of cervical cancer,and provide new treatment strategies for drug-resistant cervical cancer patients;According to three different methods,the prognosis evaluation model is constructed,and the optimal model is obtained by comparison,which provides new research ideas for further accurate medical treatment of cervical cancer patients.The experimental data used in this paper are the basic information files of cervical cancer patients downloaded from the database of the official website of Cancer Genome Map(TCGA),including RNA-seq,miRNA-seq data and clinical information data.In this paper,firstly,the standardized patient data are analyzed and calculated,and the data used in the difference analysis are the results of DESeq2 package analysis,and according to the results,thermal maps and volcanic maps are drawn to explore the molecular functions of differentially expressed genes.The specific operations of the difference analysis experiment include gene ontology(GO)analysis and Kyoto Encyclopedia of Genes and Genomes(KEGG)pathway enrichment analysis.According to the analysis of the obtained results,it is known that biological processes such as "DNA-binding transcription factor binding" and"regulation of hormone levels" are involved.In this paper,three different methods are used to construct prognosis prediction models,namely,multi-factor COX proportional hazard model,LASSO-Cox model and random forest model.The three models constructed in this paper comprehensively compare the prediction performance of cervical cancer prognosis in one year,three years and five years according to ROC values.A total of 1808 up-regulated genes and 1143 down-regulated genes were identified,totaling 2951 differentially expressed genes.The differentially expressed genes were screened.If the differential genes met the established parameter criteria,they had statistical significance and participated in the subsequent experiments.The results sho wed that a total of 152 genes were related to the survival of the patient.According to GO analysis results,the differentially expressed genes in BP(biological process)clustering were mainly concentrated in such pathways as regulation of hormone level,regulation of cell development,and proteasome protein catabolism.In CC(cell component)clustering,the protein was mainly enriched in such pathways as transcriptional cofactor activity,GTP enzyme regulatory activity,nucleoside triphosphatase regulatory activity,and signal receptor activator activity.MF clustering is mainly concentrated in actin cytoskeleton,transcription regulation complex,and cell-cell junction pathways.In the establishment of the prognostic model,the one-year,three-year and five-year AUC values of the multivariate COX proportional risk model were 0.622,0.603 and 0.613,respectively;This indicated that the constructed model was valuable for survival prediction,but it was also revealed that the prognosis prediction effect of this model was relatively poor,and further improvement was required through relevant experiments in the future.For the LASSO-Cox regression model,the AUC values in one,three,and five years are 0.76,0.84,and 0.862,respectively,indicating that the model has good prediction performance.That is to say,among the cervical cancer samples in this paper,the LASSO-Cox model is more accurate in the positive samples,and the model has better performance in the prediction of the five-year prognosis outcome of cervical cancer,so it is more likely to be used in practice.The AUC values of the random forest model in one year,three years and five years were 0.647,0.653 and 0.684,respectively,ndicating that the random forest model had certain test efficiency,and its true ability required further confirmation by relevant personnel through biological experiments.
Keywords/Search Tags:Bioinformatics, Cervical cancer, Multi-factor COX proportional risk model, LASSO regression, Random forest, Prognostic prediction
PDF Full Text Request
Related items