Conversion from snapatac2 to pycistopic

[lf96abc]

Hello,

Thank you for the package.

I am trying to convert my pre-processed snapatc2 object to pycistopic before I run topic modelling, but I'm unsure of the correct format of the input for either create_cistopic_object or create_cistopic_object_from_fragments.

In both cases, what is the correct format for path_to_fragments?
Currently, I am using the output of snpatac2.ex.export_fragments, which saves fragment files per sample in a BED format file.

For create_cistopic_object:
What is the correct format for fragment_matrix? Is it suitable to use the tile_matrix, or the peak matrix?
The tile_matrix seems wrong, as am I not using consensus peaks.

For create_cistopic_object_from_fragments:
What is the correct format for path_to_regions? Is this separated by sample?
Another user has kindly provided code for conversion by this method (#126), where path_to_regions is split by sample.
However, I have run a variation of snapatac2 peak calling without replicate across samples, so these regions are not separated by sample. Would it then be acceptable to use the same consensus regions for each sample?

In either case of create_cistopic_object or create_cistopic_object_from_fragments, is it acceptable to go straight to running the LDA modelling as shown in this tutorial?
https://github.com/aertslab/pycisTopic/blob/old/notebooks/Toy_melanoma-RTD.ipynb

I appreciate you might not be able to comment on another packages functions, but any help would be appreciated,
Thank you!