Many of our matrices are log normalized, which tends to eliminate
the number of non zero values (if there were any). This prevents
the matrix from being stored as a sparse matrix. The solution here
is to use a simple transformation to make it sparse again. The most
common value from each column is subtracted from that column. These
values that were subtracted are saved in an array called X_col_shift.
The cellxgene code needs to understand how to undo the transformation when
operating over the X matrix.
- added script to create a synthetic dataset for testing
- added a script to convert an existing CXG dataset to a sparse CXG dataset
Support for sparse tiledb arrays for the X matrix
1. cxgtool can now output sparse matrices
2. cxg_adaptor and diffexp_cxg updated to handle sparse matrices
3. added a test in test_diffexp to test sparse diffexp and get_X_array
* Improve diffexp for tiledb
- The rows from the A and B sets are gathered and processed at the same time. In this
way the matrix is only accessed once instead of twice for each tile.
- There is now a single thread queue that gets shared between all callers of the diffexp.
This will slow down work if diffexp gets too busy.
- There is a target_workunit amount of work given to each thread. Previously the
workunit was (rows selected * width of tile), which could be small. Now multiple
column tiles can be combined into one workunit. If the target is too small then
thread and other overheads may reduce performance. If target_workunit is too large
then the size of the gathered sub matrix may take up too much memory.
- add configuration parameters (max_workers, cpu_multiplier, and target_workunit)
* Specialize diffexp for tiledb
This patch adds a new diffexp algorithm which is tuned for tiledb.
This algorithm was written by Bruce and is adapted here to plug into the
current framework. The anndata_adaptor still calls the original
algotithm (which was move from diffexp.py to diffexp_generic.py).
The cxg_adaptor now calls the new diffexp_tiledb version. Some
code is shared between the two.
This is part 1 of the diffexp for tiledb. Further tuning and
global throttles are still needed.
A script to run and time diffexp with various options is also
added: test/run_diffexp.py.
* black reformat
* tighten up error reporting
* lint
* fine tuning
* additional improvements in exception handling
* lint
* include exception and traceback in log
* fix typo
* first cut at re-embedding route and back-end support
* update and expand config route tests
* add scanpy_umap
* add reembedding to config route parameters
* front-end support for reembedding fetch and UI
* remove unused imports
* add loading state
* save reembedding in reducer state
* improve withColsFrom
* transmit reembed schema to client; pick unique embedding names
* display embeddings
* format
* lint
* spaces, tab size 2
* lint
* test hack for smoke-test race
* back out hack sleep
* add check for backed mode
* add unit test for reembedding
* lint
* hide re-embedding CLI param from help
This PR contains a refactoring to make adding new features easier.
The new features include supporting the tiledb format, and the multi dataset application.
The refactoring includes
Simplifying the directory structure and files.
a class structure to handle annotations (currently one type: AnnotationsLocalFile).
a class to handle application configuration
a class structure to handle matrix data (currently AnndataAdaptor and CxgAdaptor). CxgAdaptor uses tiledb.
Algorithms that were previously dependent on the scanpy anndata object are now generalized to work with an abstract interface.
The multi dataset option is not fully supported yet, and so the option to use it is hidden.
Use "cli launch --dataroot ..."
To access this feature.
All combinations of app single dataset/ app multi dataset and AnndataAdaptor/CxgAdaptor work with all the features, such as annotations, ontologies, diffexp.
