Fig. 4

Gap statistic plot and Uniform Manifold Approximation and Projection (UMAP) of optimal clusters for the three unsupervised machine learning clustering methods: (A) hierarchical clustering, optimal number of clusters (K) is four (gap statistic = 0.52); (B) multi-omics factor analysis (MOFA) with K-means++, K is four (gap statistic = 0.43); (C) autoencoder with K-means++, K is four (gap statistic = 0.79). Error bars indicate standard deviation. The UMAPs show the distribution and shape of the identified clusters. The silhouette scores are 0.14, 0.17, and 0.52, respectively