Topacedo works with Scarf to perform memory efficient downsampling of single-cell genomics data. Using a pre-built neighbourhood graph of cells, Topacedo aims to sample cells from the graph such that the manifold is preserved. Topecedo's algorithm can be diveded into two phases:
- assigning prizes to cells and penalties to edges
- running a fast approximate prize-collecting steiner tree