Research Paper Volume 16, Issue 2 pp 1536—1554

Figure 4. Risk scores are predictive of poor survival in the TCGA-SARC set. (A) A univariate Cox analysis determined that 147 candidates from the brown module were associated with prognosis for sarcoma patients (P < 0.05). (B) The top ten genes with the highest genetic significance were selected using random forest. (C) A survival prediction model was constructed using a gene combination with a relatively small number of genes and a significant P value. (D) A higher risk score was observed in patients who died during follow-up than in people who survived. (E) A Kaplan-Meier analysis revealed poor overall survival in the high-risk score group. (F) In a principal component analysis, risk score was found to be a useful tool for evaluating a patient’s prognosis with sarcoma. (G) Overall survival was predicted well by risk scores (AUC > 0.73) in patients with sarcomas. (H) Risk score was found to be an independent risk factor for OS in both univariate and multivariate Cox regression analyses. (I) A C-index analysis concluded that the risk score had significantly more predictive power than any other clinical characteristic. Abbreviations: HR: hazard ratio; ROC: receiver operating characteristic. Statistical significance is indicated by an asterisk: ^***p < 0.001.