Fig. 5: Flexynesis can be used for unsupervised training and clustering. | Nature Communications

Fig. 5: Flexynesis can be used for unsupervised training and clustering.

From: Flexynesis: A deep learning toolkit for bulk multi-omics data integration for precision oncology and beyond

Fig. 5

The figure displays the unsupervised analysis of 21 cancer types from the TCGA study for 1600 samples (80 samples per cancer type were randomly selected). A Displays the tSNE plot of the training sample embeddings colored by the best performing clustering scheme using the k-means algorithm for values of 18 < = k < = 24, where best clustering was selected by the best silhouette score (k = 24). B The same tSNE (t-distributed Stochastic Neighbor Embedding) plot as in (A) but colored by the known cancer type labels. C The heatmap displays the concordance between the cluster labels from (A) and known cancer type labels from (B), where the adjusted mutual information score is 0.78. Each row is normalized to add up to 100% where the color of the cells represent the concordance percentage of the cancer types to the corresponding cluster labels. Source data are provided as a Source Data file.

Back to article page