Skip to main content
. 2019 Dec 30;8(12):giz153. doi: 10.1093/gigascience/giz153

Figure 2:

Figure 2:

Performance comparison of different survival analysis models in 4 cohorts. In the GPR model, we directly use the progression status as the desired output to train, whereas in the GuanRank-GPR model we convert the progression status into a complete ranking score before training. The GuanRank-GPR model (red) consistently performed best in 3 cohorts (GSE24080UAMS, HOVON65, and MMRF) and second best in the EMTAB4032 cohort, when evaluated by both (A) integrated AUC and (B) C-index. GPR, Gaussian process regression. Cox, Cox proportional hazards model. RF, random forest.