Welcome to chrysalis!
chrysalis is a powerful and lightweight method designed to identify and visualise tissue compartments in spatial
transcriptomics datasets, all without the need for external references.
chrysalis achieves this by identifying spatially variable genes (SVGs) through spatial autocorrelation.
It then employs dimensionality reduction and archetypal analysis to locate extremal points in the low-dimensional
feature space, which represent pure tissue compartments.
Each observation (i.e. capture spot) in the gene expression matrix is subsequently represented as a proportion of these
distinct compartments.
chrysalis features a unique approach based on maximum intensity projection, allowing the simultaneous visualization
of diverse tissue compartments.
Moreover, it seamlessly integrates into scanpy
based pipelines.
Discuss chrysalis on GitHub.
Get started by reading the basic tutorial.
You can also browse the API.
Consider citing our bioRxiv preprint.
Visual demonstration
human lung cancer (FFPE)
Squamous Cell Carcinoma sample by 10X Genomics.
Move the slider to reveal tissue compartments calculated by chrysalis or the associated tissue morphology.