v1.0.3

Resolves table building issue with genes containing "/" character.

v1.0.2

Duckdb version updated to most recent v1.2.2

v1.0.1

This release resolves a dependency issue when installing from v1.0.0

v1.0.0

Major release

Please see our documentation at: docs.annsql.com. Please report any bugs to the AnnSQL github issue board.

Full Preprocessing example:

docs.annsql.com/preprocessing

AnnSQL & MakeDb API

docs.annsql.com/api_annsql
docs.annsql.com/api_makedb

Extended functionality added:

• Filter by cell counts
• Filter by gene counts
• Save expression to raw layer
• Raw layer to main layer
• Save highly variable genes to main layer (X)
• Impose memory limits when instantiation of AnnSQL class
• PCA (highly experimental)
• Leiden clustering
• Umap
• Differential expression

Analysis Benchmarks added

• Filtering runtime comparisons of Seurat added (figure coming sooon)
• Filtering memory comparisons for AnnSQL, AnnData, Seurat
• Benchmark dataset generation using Splatter added for sparser filtering runtimes and memory profiles.

Considerations

• Importing h5ad files with columns > 30k. This is an issue related to the db engine we're working to mitigate. The current work around to use the make buffer file parameter in the MakeDb class.
• PCA runtime is slow; however, is memory efficient for larger datasets. We currently do not have plans to optimize this as we consider it to be highly experiment functionality. Currently, no PCA implementations exist using SQL and this is a hybrid SQL/Python approach. Additionally, the PCA method is resource intensive and will use all threads available to the system. We will release an update which limits thread usage in the near future.
• Differential expression is memory respectful and implemented as a ttest in SQL, however, there may be slower performance when comparing to other methods. This is due to the lack of matrix operation support in SQL. It will run though with limited resources, where other packages may fail.

v0.9.9.1

Extended functionality added:

• Filter by cell counts
• Filter by gene counts
• Save expression to raw layer
• Raw layer to main layer
• Save highly variable genes to main layer (X)
• Impose memory limits when instantiation of AnnSQL class
• PCA (highly experimental)

Analysis Benchmarks added

• Filtering runtime comparisons of Seurat added (figure coming sooon)
• Filtering memory comparisons for AnnSQL, AnnData, Seurat
• Benchmark dataset generation using Splatter added for sparser filtering runtimes and memory profiles.

Known Issues

• Importing h5ad files with columns > 30k. We are addressing this issue in the next release
• PCA runtime is slow; however, is memory efficient for larger datasets. We currently do not have plans to optimize this as we consider it to be highly experiment functionality. Currently, no PCA implementations exist using SQL and this is a hybrid SQL/Python approach. Additionally, the PCA method is resource intensive and will use all threads available to the system. We will release an update which limits thread usage in the near future.

Forward Functionality We will be developing extended functionality for the following below. These methods will allow users to complete a very basic full preprocessing single-cell/nuclei workflow.

• Nearest neighbors
• Leiden clustering
• Umap
• Differential expression

v0.9.8

dependency updates

v0.9.6

version v0.9.6 release

v0.9.5

Readme updates

v0.9.4

typo fixed

v0.9.3

Code cleanup