Snapatac2

Features:

- Add a argument `counting_strategy` to `pp.add_tile_matrix`, `pp.make_peak_matrix`, and `pp.make_gene_matrix`, which allows one to use different strategies
(insertion-based, fragment-based, or paired insertion counting) to compute feature counts.
- Fix 233: Add Apple silicon wheel files.

Bugs fixed:

- Fix 221: 'pp.knn' with 'method=pynndescent' invalid csr matrix.
- Fix 226: Backed AnnData does not support numpy string array.
- Fix 232: `tempdir` is not respected in `tl.macs3`.
- Fix 242: Change default value of `min_len` to `extsize` in `tl.macs3`, in order to be
consistent with the `macs3` command line tool.
- Other minor bug fixes.

Features:

- Add `count_frag_as_reads` argument to `pp.make_tile_matrix`, `pp.make_peak_matrix`,
and `pp.make_gene_matrix`.

Bugs fixed:

- Fix: `ex.export_coverage` generates empty files in v2.5.2.

Features:

- Support anndata v0.10.
- Make it easier to build custom genome objects.
- Allow customized "gene_id" and "gene_name" keys in `pp.make_gene_matrix`.
- Add `ex.export_coverage` to export coverage tracks as bedGraph or bigWig files.
This deprecates `ex.export_bigwig`.
- Add `read_10x_mtx` to read 10X mtx files.
- Implement 192: filtering fragments by size before quantification. This adds
`min_frag_size` and `max_frag_size` parameters to `pp.add_tile_matrix`,
`pp.make_peak_matrix`, and `pp.make_gene_matrix`.

Bugs fixed:

- Fix a bug in `pp.add_tile_matrix` that causes the function to fail when the genome
contains a large number (>1000) of chromosomes.
- Fix memory leak in `tl.macs3`.
- Fix 177: counter overflow in `pp.make_fragment_file` when duplicate reads are more than 2^16.
- Fix 178: issue a warning instead of an error when no cells pass the filter in `pp.import_data`.
- Fix 179: AnnDataSet to AnnData conversion error when `.X` is empty.
- Fix 182: compression type detection problem.

Function renaming:

- `ex.export_bed` is renamed to `ex.export_fragments`.

Features:

- Add zstandard file support at various places.
- Change the default output format of `ex.export_bed` to `zst`. To restore the old behavior, set `suffix='.bed.gz'`.

Bugs fixed:

- Fix a bug in `pp.import_data` that causes the function to occasionally fail on small chunk sizes.

Breaking changes:

- Fragments are now stored in `.obms['fragment_single']` or `.obsm['fragment_paired']`, depending on whether the data is single-end or paired-end. As a result, the h5ad files generated prior to this version are no longer compatible.
- `tl.call_peaks` has been renamed to `tl.macs3`, and the underlying algorithm has been significantly improved, see 142. `tl.macs3` doesn't automatically merge returned peaks anymore. Use `tl.merge_peaks` to merge peaks. Please read the updated tutorial for more details.
- `pp.import_data` now doesn't compute the TSS enrichment scores. Use `metrics.tsse` to compute the TSS enrichment scores. `genome` parameter is renamed to `chrom_sizes`.

Features:

- Add `tl.merge_peaks` to perform peak merging.
- Allow using custom mitochondrial chromosome names in `pp.import_data`.
- Change the default algorithm in `pp.knn` to kd-tree.

Bugs fixed:

- 165: Fix reproducibility issues in various functions.

Features:

- Add multiprocessing support to various preprocessing functions like `pp.import_data`.
- Add experimental `pp.import_contacts` to import scHi-C data.
- Add `pp.scanorama_integrate` to perform batch correction using Scanorama.

Bugs fixed:

- 145, 148, and others.