An Ensemble Strategy to Predict Prognosis in Ovarian Cancer Based on Gene Modules

Conclusion Considering the heterogeneity and complexity of ovarian cancer, we demonstrated a new method to predict the prognosis of ovarian cancer based on the clustering information and gene co-expression network in each subtype of cancer patients. We divided the ovarian cancer data into three subtypes by clustering analysis and we found that the survival risks in these three subtypes were significantly different. We mined the important communities based on the co-expression networks in each subtype. There are 50, 73, and 92 communities in the first, second and third subtype, respectively. Next, we constructed a new ensemble classifier based on these communities to predict the prognosis of cancer. Compared to other literature methods, our classifier had improved performance. Furthermore, the function annotation of the communities in each subtype showed that some representative communities were cancer-related and the enrichment analysis of the genes in the communities with the drug-ontology data can partially support our biomarker identification method. Author Contributions X-HZ and WZ designed the research. X-HZ and Y-CG performed the research and wrote the paper. Y-CG analyzed the data. All authors revised the manuscript. Funding This research was funded by the National Natural Science Foundation of China (61602201), the Fundamental Research Funds for the Central Universities (2662018PY023), and the National Training Program of Innovation and Entrepreneurship for Under...
Source: Frontiers in Genetics - Category: Genetics & Stem Cells Source Type: research