| Breast cancer is a kind of malignant tumor that seriously endangers women's health,It has a high degree of heterogeneity,which seriously affects individualized diagnosis and treatment.Neoadjuvant chemotherapy can reduce the risk of breast cancer recurrence,but it has no obvious use for patients who has the low risk of recurrence.Therefore it is essential to ensure reliable prognostic markers,which can provide clinical decision basis.At present,there have been studies on the analysis of breast cancer image decomposition from the perspective of tumor blood flow heterogeneity,and the prediction effect of image characteristics on prognosis has been evaluated.However,this image-based assessment method is not interpretable,and has relatively poor specificity,especially the lack of explanation of potential biological and molecular mechanisms.In contrast,gene expression data is rich in genotype and has more representative of cell functions.And from a genetic perspective,cell subpopulations can reflect the heterogeneity of cancer tissues.Therefore,this study integrates imaging data and genetic data for research,and abstracts gene expression data into a mixed weighting of cell subpopulation expressions,based on mixture convex analysis(Convex Analysis of Mixtures,CAM)decomposes different cell subpopulations from gene expression data,and establishes a correlation with prognosis according to the different proportions of the subpopulations,and explains it from the perspective of molecular function,and then obtains The conclusions were combined with imaging methods to find prognostic image markers.The specific research contents of this thesis include:(1)Decomposition of gene expression data: Decomposing breast cancer gene expression data based on the CAM model,determining the number of decomposed subpopulations by the minimum description length,and then we obtain the subpopulation proportion matrix,specific expression matrix,and marker genes.And the product of the two matrices obtained by the decomposition is approximately equivalent to the original gene expression data.The analysis of marker gene pathways reveals the biological significance and molecular mechanism of each subpopulation.(2)Study on the association between the subgroups generated by decomposition and prognosis: The subpopulation proportion matrix reflects the proportion of gene expression information in each subpopulation,which is the main reason for the difference in patient expression data.Therefore,survival analysis was performed on the proportion information of each subpopulation to explore its relationship with prognosis.Then cluster the characteristic subpopulation related to prognosis to obtain subpopulation subtypes,and the association with prognosis is established.Finally,the prognostic value of subpopulation subtypes is verified in an independent test set.(3)Prognosis analysis after the conclusion of gene expression analysis based on the method of image genomics mapping to the image: matching gene expression data and image data,and establishing the relationship between gene expression and image features,and then establishing the logistic regression classification prediction model,we predict the patient labeling in the independent verification set.The results showed that there were survival differences among different types of patients.Through studying the correlation between the gene expression data and prognosis of breast cancer in this paper,it is proved that subpopulation can play a role in influencing prognosis,and its corresponding marker gene has the potential to be a biomarker of prognostic molecule.Meanwhile,through studying the correlation between gene expression data and image characteristics,the latent imaging prognostic markers are founded.Maybe the molecular markers and image markers determined by this study can provide valid information of prognostic prediction for breast cancer patients and also provide theoretical basis and treatment target for precise treatment. |
PDF Full Text Request