Skip to main content
Fig. 7 | BMC Bioinformatics

Fig. 7

From: Contrastive self-supervised clustering of scRNA-seq data

Fig. 7

Visualization of identified clusters. scDeepCluster, scziDesk, desc, contrastive + KM, contrastive + LD identified clusters on 4 datasets (Quake Limb Muscle, Muraro, Romanov, Young) are compared using a 2D data projection. All selected methods start by creating an embedding for the cells which is clustered in a second phase. The quality of the method depends on both the created embedding and the clustering algorithm. For scDeepCluster and scziDesk, we depicted the clustering prediction relative to the underlying embedding. Our methods (contrastive + KM and contrastive + LD) clustered the same embedding. All plots present a 2D t-SNE projection of the underlying embeddings

Back to article page