How to Embed BBKNN Results into MultimodalData?

[x1han]

Hello Developers,

We are currently working on the integration of single-cell data samples and embedding them into Pegasus's MultimodalData. For Seurat's CCA, we have created a separate UnimodalData from the generated matrix and embed the integrated PCA results into obsm. However, we are having some difficulties with scanpy's BBKNN (which often produces good results) because the way scanpy and Pegasus store the results of the neighbors function seems to differ.

Therefore, my question is: If I generate BBKNN results in scanpy ('neighbors' in adata.uns, 'distances' in adata.obsp, 'connectivities' in adata.obsp), which results need to be transferred into Pegasus to ensure that subsequent dimensionality reduction and clustering can proceed normally?