Performance & Scalability

Cellestial builds a polars.DataFrame from the single-cell data type (e.g. AnnData) to be passed into ggplot(), either directly or after necessary processing. The process of building the DataFrame mostly avoids using unnecessary columns allowing minimal memory usage and scalability. Leveraging the blazing fast nature of Polars, Cellestial ensures performance.

Plotting, the core step, is handled by Lets-Plot which is fast and scalable.

Example: 100k cells

Using Human Breast Cancer Single Cell Atlas.