Fig. 3

Survival analysis using combinations of different data types and benchmark experiment on ten cancer types. A Up: Survival differences comparison using different omics combination for can subtyping. Down: Weight values for each data type. B Boxplot shows the survival performance of the nine integration methods on ten cancer datasets. MOSD outperforms other eight methods indicated by the -log10(p-value). Meanwhile, our method exhibits impressive computational efficiency (C) and relatively better silhouette width (D) among the nine methods. Bars indicate the average performance over the ten cancer types. Error bars is the 95% confidence interval