Snapatac2

Breaking changes:

- Due to a major upgrade of the `anndata-rs` package. The `AnnData` object generated by older versions is no longer compatible with the new version.
- All dataframes in backed mode are now indexless.
- The `pp.call_doublets` and `pl.scrublet` has been removed. `pp.scrublet` now automatically calls doublets.
- Rename the argument `gff_file` to `gene_anno` in `pp.import_data` and `pp.make_gene_matrix`. Because the gene annotation file is no longer required to be a GFF file.

Features:

- Updates to `tl.spectral`:
1. Change the default similarity metric to the cosine similarity.
2. Significantly improve the scalability of the algorithm when using the cosine similarity. Both the time and space complexity are now linear!
3. The returned eigenvectors are now weighted by eigenvalues so that using the elbow method to select the number of eigenvectors is no longer necessary. To restore the old behavior, set `weighted_by_sd=False`.
- Implement `tl.multi_spectral` to perform dimension reduction on multiple modalities simultaneously.
- Most preprocessing functions can now take a list of `AnnData` objects as input, and process them in parallel.
- `pp.make_tile_matrix`: allow excluding certain chromosomes by setting `exclude_chroms`.

Bugs fixed:

- Fix a bug in `pp.call_doublets` that leads to incorrect thresholds for certain datasets.
- Fix 80, 97, 102, 103, 109, 110.

Breaking changes:

- `create_dataset` has been removed. Please use the `AnnDataSet` constructor to create the AnnDataSet object.
- Argument `storage` is renamed to `file` in a variety of AnnData IO functions, including `create_dataset`, `read_mtx`, and `read_csv`.
- Argument `no_check` is removed in `read_dataset`.
- `mode` is renamed to `backed` in `.read()`.

Features:

- Support in-memory AnnData and flexible conversions between backed and in-memory AnnData.

Features:

- Add a low-memory mode in `pp.import_data` for parsing large unsorted fragment files.

Bugs fixed:

- Minor bug fixes and performance improvements.
- Fix 10X bam file parsing in `pp.make_fragment_file`.

Bugs fixed:

- Minor bug fixes and performance improvements.

Features:

- Add "tl.motif_enrichment" to perform motif enrichment analysis.

Bugs fixed:

- Fix bugs in "ex.export_bigwig". NOTE: prior to this version, this function does not work as expected. Please update and rerun if you have used this function in 2.1.0.

Breaking changes:

- "pp.make_tile_matrix" has been renamed to "pp.add_tile_matrix".
- "snapatac2.genome" has been redesigned. A new "Genome" class is added for automatic download of gene annotation files. Objects in "snap.genome.*" are now instances of the Genome object.
- Rename each occurrence of "group_by" to "groupby" to be consistent with other packages.

Features:

- Add "tl.diff_test" to identify differentially accessible regions.
- Add "tl.make_fragment_file" to convert BAM files to fragment files.
- Add various functions to perform transcriptional regulatory network analysis.
- Add "read_10x_mtx" for reading files produced by 10X CellRanger pipeline.
- Add "ex.export_bigwig" to generate bigwig files.

Bugs fixed:

- Various bug fixes and performance improvements.