Extended Data Fig. 9: Strategy and model performance of the integrative modelling on multi-omics associated with anti-PD-1+acarbose combination therapy.
a, Swim lane plots for classification analysis (n = 5 per group). Each swim lane presented predictions for one mice sample. The predictions of class probabilities were colour coded by class and presented per repetition (smaller dots) and averaged over all repetitions (larger dots). And the misclassified predictions were represented as circled dots. b, Multi-omics correlation plot indicating the correlations between the first principal component from each omics data and the power to separate sample from distinct groups. c,The circus plot and relevance network demonstrated the either positive (red lines) or negative (blue lines) correlations between the selected multi-omics features (n = 5 per group). The additional line plots on the outer ring visualized the relative abundance of each feature per group. The size of each node in the relevance network represents the betweenness centrality. d, The clustered heatmap (Euclidean distance, complete linkage) represented the scaled expression of the selected multi-omics signatures (n = 5 per group). Samples were represented in rows, selected features on the first component in columns. The red and the blue coloured blocks represented a positive or a negative association respectively, while the yellow colour represented weak levels of association.
