Fig. 3: Ligand separation and feature distribution in full MCF10A dataset. | Communications Biology

Fig. 3: Ligand separation and feature distribution in full MCF10A dataset.

From: A multi-encoder variational autoencoder controls multiple transformational features in single-cell image analysis

Fig. 3

a UMAP embeddings for respective VAE encodings, allowing for qualitative visual evaluation of ligand separability. Cluster purities and normalized mutual information were calculated to quantitatively compare methods (k-means clusters = 12 to allow for ligand subpopulations). The mean cluster purity of the standard VAE was 0.04 with a standard deviation of 0.03. The mean cluster purity of the ME-VAE was 0.59 with a standard deviation of 0.15. The total sample size is n = 73,134 single-cell images. b Distribution of stain features across UMAP space, colored by intensity.

Back to article page