Anndata integration (#60)

* removed plotting * cunnData deprecated * added anndata and 64 bit support * switch from scanpy_gpu to tools * start to updating docs * update test to work with anndata * swtiched harmony to pp * switched harmony to pp * updated docs * fix docs * added utils * updated documentation * updated docs * move to anndata * updated doc strings * use next anndata * adata version * updated docs * try to fix utils docs * update for dense hvg * added dense normalization test * update and test utils * update test * Update docs/Usage_Principles.md Co-authored-by: Lukas Heumos <[email protected]> * Update docs/Usage_Principles.md Co-authored-by: Lukas Heumos <[email protected]> * Update docs/Usage_Principles.md Co-authored-by: Lukas Heumos <[email protected]> * Update docs/Usage_Principles.md Co-authored-by: Lukas Heumos <[email protected]> * Update docs/Usage_Principles.md Co-authored-by: Lukas Heumos <[email protected]> * Update docs/Usage_Principles.md Co-authored-by: Lukas Heumos <[email protected]> * update api docs * updated docs * test docs * fix docs * added release-notes * updated 0.8.1 * fix for realease-notedocs * fixed links * try to fix release-notes * bumped scanpydoc * test css * make c array safe * enables cunndata to anndata_GPU * update notebooks * updated docs * maybe fix docs * update utils docstring * reset docs * added extlinks * fixes AnnData * update AnnData * added AnnData * updated docs * fixes broken link * test releasenotes * updated conf.py * creates utils in docs * renamed Params Parameters * switched to boolean index * added batching * prepare for transformer API * enhance harmony reproducability * harmony 64 bit * work with dense matrices * added release node * updated ruff * update notebooks * remove cpx raw import * remove cugraph default import * fixed import * make doc-strings pretty * added leiden disclaimer to notebooks * added test for harmonypy * added import test * added profimp for testing * updated anndata-dep --------- Co-authored-by: Lukas Heumos <[email protected]> Co-authored-by: Philipp A <[email protected]>
scverse · Sep 9, 2023 · ba36e67 · ba36e67
1 parent b65e766
commit ba36e67
Show file tree

Hide file tree

Showing 77 changed files with 3,916 additions and 3,663 deletions.
diff --git a/.gitignore b/.gitignore
@@ -12,6 +12,7 @@ __pycache__/
 # Sphinx documentation
 /docs/_build/
 /docs/generated/
+/docs/api/generated/
 
 # Venvs
 *venv/
diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
@@ -14,7 +14,7 @@ repos:
     hooks:
     -   id: black
 -   repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: v0.0.286
+    rev: v0.0.287
     hooks:
     -   id: ruff
         args: [--fix, --exit-non-zero-on-fix]
diff --git a/docs/Usage_Principles.md b/docs/Usage_Principles.md
@@ -6,36 +6,46 @@
 import rapids_singlecell as rsc
 ```
 
-## cunnData
+## Workflow
 
-```{image} _static/cunndata.svg
-:width: 500px
-```
+The workflow of *rapids-singlecell* is basically the same as *scanpy's*. The main difference is the speed at which *rsc* can analyze the data. For more information please checkout the notebooks and API documentation.
 
-The {class}`~rapids_singlecell.cunnData.cunnData` object replaces {class}`~anndata.AnnData` for preprocessing. All {mod}`~.pp` and {mod}`~.pl` functions are aimed towards cunnData. {attr}`~.X` and {attr}`~.layers` are stored on the GPU. The other components are stored in the host memory.
+### AnnData setup
 
-## Workflow
+With the release of version 0.10.0 {class}`~anndata.AnnData` supports GPU arrays and Sparse Matrices.
 
-The workflow of *rapids-singlecell* is basically the same as *scanpy's*. The main difference is the speed at which *rsc* can analyze the data. For more information please checkout the notebooks and API documentation.
+Rapids-singlecell leverages this capability to perform analyses directly on GPU-enabled {class}`~anndata.AnnData` objects. This also leads to the depreciation of {class}`~rapids_singlecell.cunnData.cunnData` and it's removal in early 2024.
+
+To get your {class}`~anndata.AnnData` object onto the GPU you can set {attr}`~anndata.AnnData.X` or each {attr}`~anndata.AnnData.layers` to a GPU based matrix.
+
+```
+adata.X = cpx.scipy.sparse.csr_matrix(adata.X)  # moves `.X` to the GPU
+adata.X = adata.X.get() # moves `.X` back to the CPU
+```
+
+You can also use {mod}`rapids_singlecell.utils` to move arrays and matrices.
+
+```
+rsc.utils.anndata_to_GPU(adata) # moves `.X` to the GPU
+rsc.utils.anndata_to_CPU(adata) # moves `.X` to the CPU
+```
 
 ### Preprocessing
 
-The preprocessing is handled by {class}`~rapids_singlecell.cunnData.cunnData` and `cunnData_funcs`. The latter is import as {mod}`~.pp` and {mod}`~.pl` to mimic the behavior of scanpy.
+The preprocessing can  be handled by {class}`~anndata.AnnData` and {class}`~rapids_singlecell.cunnData.cunnData`. It offers accelerated versions of functions within {mod}`scanpy.pp`.
 
 Example:
 ```
-rsc.pp.highly_variable_genes(cudata,n_top_genes=5000,flavor="seurat_v3",batch_key= "PatientNumber",layer = "counts")
-cudata = cudata[:,cudata.var["highly_variable"]==True]
-rsc.pp.regress_out(cudata,keys=["n_counts", "percent_MT"])
-rsc.pp.scale(cudata,max_value=10)
+rsc.pp.highly_variable_genes(adata, n_top_genes=5000, flavor="seurat_v3", batch_key= "PatientNumber", layer = "counts")
+adata = adata[:,adata.var["highly_variable"]==True]
+rsc.pp.regress_out(adata,keys=["n_counts", "percent_MT"])
+rsc.pp.scale(adata,max_value=10)
 ```
 After preprocessing is done just transform the {class}`~rapids_singlecell.cunnData.cunnData` into {class}`~anndata.AnnData` and continue the analysis.
 
 ### Tools
 
-The functions provided in {mod}`~.tl` are designed to manipulate the {class}`~anndata.AnnData` object. They serve as near drop-in replacements for the functions in *scanpy*, but offer significantly improved performance. Consequently, you can continue to use scanpy's plotting API with ease.
-
-All {mod}`~.tl` functions operate on {class}`~anndata.AnnData`, which is why {func}`~.harmony_integrate` has transitioned from the `.pp` module to `.tl`. This also explains the existence of two distinct functions for calculating principal components: one for {class}`~rapids_singlecell.cunnData.cunnData` (within the `.pp` module) and another for {class}`~anndata.AnnData` (within the `.tl` module).
+The functions provided in {mod}`~.tl` are designed to as near drop-in replacements for the functions in {mod}`scanpy.tl`, but offer significantly improved performance. Consequently, you can continue to use scanpy's plotting API.
 
 Example:
 ```
@@ -55,3 +65,11 @@ rsc.dcg.run_mlm(mat=adata, net=net, source='source', target='target', weight='we
 acts_mlm = dc.get_acts(adata, obsm_key='mlm_estimate')
 sc.pl.umap(acts_mlm, color=['KLF5',"FOXA1", 'CellType'], cmap='coolwarm', vcenter=0)
 ```
+
+### cunnData (deprecated)
+
+```{image} _static/cunndata.svg
+:width: 500px
+```
+
+The {class}`~rapids_singlecell.cunnData.cunnData` object can replace {class}`~anndata.AnnData` for preprocessing. All {mod}`~.pp` functions (except {func}`~.pp.neighbors`) are aimed towards cunnData. {attr}`~.X` and {attr}`~.layers` are stored on the GPU. The other components are stored in the host memory.
diff --git a/docs/_static/css/override.css b/docs/_static/css/override.css
@@ -0,0 +1,20 @@
+/* for the sphinx design cards */
+body {
+    --sd-color-shadow: dimgrey;
+}
+
+dl.citation>dt {
+    float: left;
+    margin-right: 15px;
+    font-weight: bold;
+}
+
+/* for custom small role */
+.small {
+    font-size: 40% !important
+}
+
+.smaller,
+.pr {
+    font-size: 70% !important
+}
diff --git a/docs/api.md b/docs/api.md
diff --git a/docs/api/cunndata.md b/docs/api/cunndata.md
@@ -0,0 +1,13 @@
+# cunnData
+
+{class}`~rapids_singlecell.cunnData.cunnData` is deprecated and will be removed in 2024. Please start switching to {class}`~anndata.AnnData`
+
+```{eval-rst}
+.. module:: rapids_singlecell.cunnData
+.. currentmodule:: rapids_singlecell
+
+.. autosummary::
+    :toctree: generated
+
+    cunnData.cunnData
+```
diff --git a/docs/api/decoupler_gpu.md b/docs/api/decoupler_gpu.md
@@ -0,0 +1,14 @@
+# decoupler-GPU: `dcg`
+
+{mod}`decoupler` contains different statistical methods to extract biological activities. {mod}`rapids_singlecell.dcg` acclerates some of these methods.
+
+```{eval-rst}
+.. module:: rapids_singlecell.dcg
+.. currentmodule:: rapids_singlecell
+
+.. autosummary::
+    :toctree: generated
+
+    dcg.run_mlm
+    dcg.run_wsum
+```
diff --git a/docs/api/index.md b/docs/api/index.md
@@ -0,0 +1,18 @@
+# API
+
+Import rapids-singlecell as:
+
+```
+import rapids_singlecell as rsc
+```
+
+
+```{toctree}
+:maxdepth: 2
+
+scanpy_gpu
+squidpy_gpu
+decoupler_gpu
+utils
+cunndata
+```