annotate dimension_reduction_clustering.xml @ 6:d258720d9a42 draft default tip

planemo upload for repository https://github.com/galaxyproject/tools-iuc/tree/main/tools/snapatac2 commit e0f59fae19e57f54ae0c351a16dd1805d12aba1d

<tool id="snapatac2_clustering" name="SnapATAC2 Clustering" version="@TOOL_VERSION@+galaxy@VERSION_SUFFIX@" profile="@PROFILE@">
    <description>and dimension reduction</description>
    <macros>
        <import>macros.xml</import>
    </macros>
    <expand macro="xrefs"/>
    <requirements>
        <expand macro="requirements"/>
    </requirements>
    <command detect_errors="exit_code"><![CDATA[
export NUMBA_CACHE_DIR="\${TEMP:-/tmp}";
#if $method.method == 'tl.multi_spectral'
    #for $i in range(len($method.adata))
        cp $method.adata[$i] 'adata_${i}.h5ad' &&
    #end for
#else
    @CMD_PREP_ADATA@
#end if
@CMD@
    ]]></command>
    <configfiles>
        <configfile name="script_file"><![CDATA[

@CONF_IMPORTS@
#if $method.method == 'tl.multi_spectral'
## read all files ending with .h5ad in the working directory
import glob
files = sorted(glob.glob('adata_*.h5ad'))

adata_list = []
for fn in files:
    ad = snap.read(fn, backed=None)
    adata_list.append(ad)
#else
@CONF_READ_INPUTS@
#end if

#if $method.method == 'tl.spectral'
    #if $method.features
with open('$method.features') as f:
    features_mask = [x.lower().capitalize() == "True" for x in f.read().splitlines()]
    #end if

## somewhere in the SnapATAC2 code, a pandas Series is being passed where a numpy array is expected.
## This is a workaround to add the nonzero method back to pandas Series.
## Add the nonzero method back to pandas Series
import pandas as pd
def series_nonzero(self):
    return (self != 0).values.nonzero()

pd.Series.nonzero = series_nonzero


snap.tl.spectral(
    adata,
    n_comps = $method.n_comps,
    #if $method.features
    features = features_mask,
    #end if
    random_state = $method.random_state,
    #if $method.sample_size
    sample_size = $method.sample_size,
    #end if
    chunk_size = $method.chunk_size,
    distance_metric = '$method.distance_metric',
    weighted_by_sd = $method.weighted_by_sd,
    inplace = True
)

#else if $method.method == 'tl.multi_spectral'
    #if $method.features
with open('$method.features') as f:
    features_mask = [x.lower().capitalize() == "True" for x in f.read().splitlines()]
    #end if

embedding = snap.tl.multi_spectral(
    adatas = adata_list,
    n_comps = $method.n_comps,
    #if $method.features
    features = features_mask,
    #else
    features = None,
    #end if
    weights = None,  # Will enable if requested by users
    random_state = $method.random_state,
    weighted_by_sd = $method.weighted_by_sd,
)

adata = adata_list[0].copy()
adata.uns['spectral_eigenvalue_joint'] = embedding[0]
adata.obsm['X_joint'] = embedding[1]

#else if $method.method == 'tl.umap'
snap.tl.umap(
    adata,
    n_comps = $method.n_comps,
    #if $method.use_dims != ''
        #set $dims = ([x.strip() for x in str($method.use_dims).split(',')])
    use_dims=$dims,
    #end if
    use_rep = '$method.use_rep',
    key_added = '$method.key_added',
    random_state = $method.random_state,
    inplace = True
)

#else if $method.method == 'pp.knn'
snap.pp.knn(
    adata,
    n_neighbors = $method.n_neighbors,
    #if $method.use_dims != ''
        #set $dims = ([x.strip() for x in str($method.use_dims).split(',')])
    use_dims=$dims,
    #end if
    use_rep = '$method.use_rep',
    method = '$method.algorithm',
    inplace = True,
    random_state = $method.random_state
)

#else if $method.method == 'tl.leiden'
snap.tl.leiden(
    adata,
    resolution = $method.resolution,
    objective_function = '$method.objective_function',
    min_cluster_size = $method.min_cluster_size,
    n_iterations = $method.n_iterations,
    random_state = $method.random_state,
    key_added = '$method.key_added',
    use_leidenalg = $method.use_leidenalg,
    weighted = $method.weighted,
    inplace = True
)

#else if $method.method == 'tl.kmeans'
snap.tl.kmeans(
    adata,
    n_clusters = $method.n_clusters,
    n_iterations = $method.n_iterations,
    random_state = $method.random_state,
    use_rep = '$method.use_rep',
    key_added = '$method.key_added'
)

#else if $method.method == 'tl.dbscan'
snap.tl.dbscan(
    adata,
    eps = $method.eps,
    min_samples = $method.min_samples,
    leaf_size = $method.leaf_size,
    use_rep = '$method.use_rep',
    key_added = '$method.key_added',
    n_jobs = int(os.getenv("GALAXY_SLOTS", 4))
)

#else if $method.method == 'tl.hdbscan'
snap.tl.hdbscan(
    adata,
    min_cluster_size = $method.min_cluster_size,
    #if $method.min_samples
    min_samples = $method.min_samples,
    #end if
    cluster_selection_epsilon = $method.cluster_selection_epsilon,
    alpha = $method.alpha,
    cluster_selection_method = '$method.cluster_selection_method',
    random_state = $method.random_state,
    use_rep = '$method.use_rep',
    key_added = '$method.key_added'
)

## It is implemented in select_feature function, and it is problematic if the user don't select a groupby  (will return an array). i think this can be skipped unless needed
## #else if $method.method == 'tl.aggregate_X'
## snap.tl.aggregate_X(
##     adata,
##     #if $method.groupby != ''
##     groupby = '$method.groupby',
##     #end if
##     normalize = '$method.normalize'
## )

#else if $method.method == 'tl.aggregate_cells'
snap.tl.aggregate_cells(
    adata,
    use_rep = '$method.use_rep',
    #if $method.target_num_cells
    target_num_cells = $method.target_num_cells,
    #end if
    min_cluster_size = $method.min_cluster_size,
    random_state = $method.random_state,
    key_added = '$method.key_added',
    inplace = True
)
#end if

@CONF_ANNDATA_WRITE_OUTPUTS@
    ]]></configfile>
    </configfiles>
    <inputs>
        <conditional name="method">
            <param name="method" type="select" label="Dimension reduction and Clustering">
                <option value="tl.spectral">Perform dimension reduction using Laplacian Eigenmap, using 'tl.spectral'</option>
                <option value="tl.multi_spectral">similar to 'tl.multi_spectral' but it can work on multiple modalities.</option>
                <option value="tl.umap">Compute Umap, using 'tl.umap'</option>
                <option value="pp.knn">Compute a neighborhood graph of observations, using 'pp.knn'</option>
                <option value="tl.leiden">Cluster cells into subgroups, using 'tl.leiden'</option>
                <option value="tl.kmeans">Cluster cells into subgroups using the K-means algorithm, using 'tl.kmeans'</option>
                <option value="tl.dbscan">Cluster cells into subgroups using the DBSCAN algorithm, using 'tl.dbscan'</option>
                <option value="tl.hdbscan">Cluster cells into subgroups using the HDBSCAN algorithm, using 'tl.hdbscan'</option>
                <!-- It is implemented in select_feature function in preprocessing.xml tool (implemented in upstream code, not in the xml). It is problematic if the user don't select a groupby  (will return an array). i think this can be skipped unless needed -->
                <!-- <option value="tl.aggregate_X">Aggregate values in adata.X in a row-wise fashion, using 'tl.aggregate_X'</option> -->
                <option value="tl.aggregate_cells">Aggregate cells into pseudo-cells, using 'tl.aggregate_cells'</option>
            </param>
            <when value="tl.spectral">
                <expand macro="param_inputs_anndata"/>
                <expand macro="param_n_comps"/>
                <param argument="features" type="data" format="txt,tabular" optional="true" label="Text file indicating features to keep. Each line contains only word (True/False)." help="True means that the feature is kept. False means the feature is removed"/>
                <expand macro="param_random_state"/>
                <param argument="sample_size" type="float" min="0" max="1" optional="true" label="Approximate the embedding using the Nystrom algorithm by selecting a subset of cells" help="Using this only when the number of cells is too large, e.g. &gt; 10,000,000, or the `distance_metric` is “jaccard”"/>
                <param argument="chunk_size" type="integer" value="20000" label="chunk size"/>
                <param argument="distance_metric" type="select" label="distance metric: “jaccard”, “cosine“">
                    <option value="jaccard">jaccard</option>
                    <option value="cosine" selected="true">cosine</option>
                </param>
                <param argument="weighted_by_sd" type="boolean" truevalue="True" falsevalue="False" checked="true" label="Whether to weight the result eigenvectors by the square root of eigenvalues"/>
            </when>
            <when value="tl.multi_spectral">
                <expand macro="param_inputs_anndata" multiple="true" label="list of Anndatas to use for multi_spectral" help="Please note that the embedding will be saved in the first Anndata"/>
                <!-- Will enable if requested by users -->
                <!-- <param name="weights" type="data" format="tabular" optional="true" label="Weights" help="Text file indicating weights for each modality. Each line contains only floats.If not provided, all modalities are weighted equally" /> -->
                <expand macro="param_n_comps"/>
                <param argument="features" type="data" format="txt,tabular" optional="true" label="Text file indicating features to keep. Each line contains only word (True/False)." help="True means that the feature is kept. False means the feature is removed"/>
                <expand macro="param_random_state"/>
                <param argument="weighted_by_sd" type="boolean" truevalue="True" falsevalue="False" checked="true" label="Whether to weight the result eigenvectors by the square root of eigenvalues"/>
            </when>
            <when value="tl.umap">
                <expand macro="param_inputs_anndata"/>
                <expand macro="param_n_comps" value="2" label="Number of components" help=""/>
                <param argument="use_dims" type="text" optional="true" label="Use these dimensions in `use_rep`" help="comma separated list of dimensions">
                    <expand macro="sanitize_query"/>
                </param>
                <expand macro="param_use_rep"/>
                <expand macro="param_key_added" key_added="umap"/>
                <expand macro="param_random_state"/>
            </when>
            <when value="pp.knn">
                <expand macro="param_inputs_anndata"/>
                <param argument="n_neighbors" type="integer" value="50" label="The number of nearest neighbors to be searched"/>
                <param argument="use_dims" type="text" value="" optional="true" label="The dimensions used for computation">
                    <expand macro="sanitize_query"/>
                </param>
                <expand macro="param_use_rep" label="The key for the matrix"/>
                <param argument="algorithm" type="select" label="Choose method">
                    <option value="kdtree" selected="true">'kdtree': use the kdtree algorithm to find the nearest neighbors</option>
                    <option value="hora">'hora': use the HNSW algorithm to find the approximate nearest neighbors</option>
                    <option value="pynndescent">'pynndescent': use the pynndescent algorithm to find the approximate nearest neighbors</option>
                </param>
                <expand macro="param_random_state" label="Random seed for approximate nearest neighbor search" help="Note that this is only used when method='pynndescent'. Currently hora does not support random seed, so the result of hora is not reproducible."/>
            </when>
            <when value="tl.leiden">
                <expand macro="param_inputs_anndata"/>
                <param argument="resolution" type="float" value="1" label="Parameter value controlling the coarseness of the clustering" help="Higher values lead to more clusters"/>
                <param argument="objective_function" type="select" label="Whether to use the Constant Potts Model (CPM) or modularity">
                    <option value="CPM">CPM</option>
                    <option value="modularity" selected="true">modularity</option>
                    <option value="RBConfiguration">RBConfiguration</option>
                </param>
                <param argument="min_cluster_size" type="integer" value="5" label="The minimum size of clusters"/>
                <expand macro="param_n_iterations"/>
                <expand macro="param_random_state"/>
                <expand macro="param_key_added" key_added="leiden"/>
                <param argument="use_leidenalg" type="boolean" truevalue="True" falsevalue="False" checked="false" label="Whether to use the leidenalg package for Leiden clustering"/>
                <param argument="weighted" type="boolean" truevalue="True" falsevalue="False" label="Whether to use the edge weights in the graph"/>
            </when>
            <when value="tl.kmeans">
                <expand macro="param_inputs_anndata"/>
                <param argument="n_clusters" type="integer" value="5" label="Number of clusters to return"/>
                <expand macro="param_n_iterations"/>
                <expand macro="param_random_state"/>
                <expand macro="param_use_rep"/>
                <expand macro="param_key_added" key_added="kmeans"/>
            </when>
            <when value="tl.dbscan">
                <expand macro="param_inputs_anndata"/>
                <param argument="eps" type="float" value="0.5" label=" The maximum distance between two samples for one to be considered as in the neighborhood of the other" help="This is not a maximum bound on the distances of points within a cluster. This is the most important DBSCAN parameter to choose appropriately for your data set and distance function."/>
                <param argument="min_samples" type="integer" value="5" label="The number of samples (or total weight) in a neighborhood for a point to be considered as a core point."/>
                <param argument="leaf_size" type="integer" value="30" label="Leaf size passed to BallTree or cKDTree" help="This can affect the speed of the construction and query, as well as the memory required to store the tree."/>
                <expand macro="param_use_rep"/>
                <expand macro="param_key_added" key_added="dbscan"/>
            </when>
            <when value="tl.hdbscan">
                <expand macro="param_inputs_anndata"/>
                <param argument="min_cluster_size" type="integer" value="5" label="The minimum size of clusters"/>
                <param argument="min_samples" type="integer" value="" optional="true" label="The number of samples in a neighborhood for a point to be considered a core point"/>
                <param argument="cluster_selection_epsilon" type="float" value="0.0" label="A distance threshold. Clusters below this value will be merged"/>
                <param argument="alpha" type="float" value="1.0" label="A distance scaling parameter as used in robust single linkage"/>
                <param argument="cluster_selection_method" type="select" label="The method used to select clusters from the condensed tree">
                    <option value="eom">Excess of Mass algorithm to find the most persistent clusters</option>
                    <option value="leaf">Select the clusters at the leaves of the tree - this provides the most fine grained and homogeneous clusters</option>
                </param>
                <expand macro="param_random_state"/>
                <expand macro="param_use_rep"/>
                <expand macro="param_key_added" key_added="hdbscan"/>
            </when>
            <!-- <when value="tl.aggregate_X">
                <expand macro="param_inputs_anndata"/>
                <expand macro="param_groupby"/>
                <param argument="normalize" type="select" optional="true" label="normalization method">
                    <option value="RPM">RPM</option>
                    <option value="RPKM">RPKM</option>
                </param>
            </when> -->
            <when value="tl.aggregate_cells">
                <expand macro="param_inputs_anndata"/>
                <expand macro="param_use_rep"/>
                <param argument="target_num_cells" type="integer" value="" optional="true" label="target_num_cells" help="If None, `target_num_cells = num_cells / min_cluster_size`"/>
                <param argument="min_cluster_size" type="integer" value="50" label="The minimum size of clusters"/>
                <expand macro="param_random_state"/>
                <expand macro="param_key_added" key_added="pseudo_cell"/>
            </when>
        </conditional>
        <expand macro="param_common_advanced"/>
    </inputs>
    <outputs>
        <data name="anndata_out" format="h5ad" from_work_dir="anndata.h5ad.gz" label="${tool.name} (${method.method}) on ${on_string}: Annotated data matrix"/>
        <data name="hidden_output" format="txt" label="Log file">
            <filter>advanced_common['show_log']</filter>
        </data>
    </outputs>
    <tests>
        <test expect_num_outputs="2">
            <!-- tl.spectral -->
            <conditional name="method">
                <param name="method" value="tl.spectral"/>
                <param name="adata" location="https://zenodo.org/records/17512085/files/pp.select_features.pbmc_500_chr21.h5ad"/>
                <param name="n_comps" value="30"/>
                <param name="random_state" value="0"/>
                <param name="chunk_size" value="20000"/>
                <param name="distance_metric" value="jaccard"/>
                <param name="weighted_by_sd" value="True"/>
            </conditional>
            <section name="advanced_common">
                <param name="show_log" value="true"/>
            </section>
            <output name="hidden_output">
                <assert_contents>
                    <has_text_matching expression="snap.tl.spectral"/>
                    <has_text_matching expression="random_state = 0"/>
                    <has_text_matching expression="n_comps = 30"/>
                    <has_text_matching expression="chunk_size = 20000"/>
                    <has_text_matching expression="distance_metric = 'jaccard'"/>
                    <has_text_matching expression="weighted_by_sd = True"/>
                </assert_contents>
            </output>
            <output name="anndata_out" ftype="h5ad">
                <assert_contents>
                    <has_h5_keys keys="uns/spectral_eigenvalue"/>
                    <has_h5_keys keys="obsm/X_spectral"/>
                </assert_contents>
            </output>
        </test>
        <test expect_num_outputs="2">
            <!-- tl.multi_spectral -->
            <conditional name="method">
                <param name="method" value="tl.multi_spectral"/>
                <param name="adata" location="https://zenodo.org/records/17512085/files/snap_datasets_pbmc10k_multiome_RNA.h5ad.gz,https://zenodo.org/records/17512085/files/snap_datasets_pbmc10k_multiome_ATAC.h5ad.gz"/>
                <param name="n_comps" value="30"/>
                <param name="random_state" value="0"/>
                <param name="weighted_by_sd" value="True"/>
            </conditional>
            <section name="advanced_common">
                <param name="show_log" value="true"/>
            </section>
            <output name="hidden_output">
                <assert_contents>
                    <has_text_matching expression="snap.tl.multi_spectral"/>
                    <has_text_matching expression="random_state = 0"/>
                    <has_text_matching expression="n_comps = 30"/>
                    <has_text_matching expression="weighted_by_sd = True"/>
                    <has_text_matching expression="features = None"/>
                </assert_contents>
            </output>
            <output name="anndata_out" ftype="h5ad">
                <assert_contents>
                    <has_h5_keys keys="uns/spectral_eigenvalue_joint"/>
                    <has_h5_keys keys="obsm/X_joint"/>
                </assert_contents>
            </output>
        </test>
        <test expect_num_outputs="2">
            <!-- tl.umap -->
            <conditional name="method">
                <param name="method" value="tl.umap"/>
                <param name="adata" location="https://zenodo.org/records/17512085/files/tl.spectral.pbmc_500_chr21.h5ad"/>
                <param name="n_comps" value="2"/>
                <param name="use_rep" value="X_spectral"/>
                <param name="key_added" value="umap"/>
                <param name="random_state" value="0"/>
            </conditional>
            <section name="advanced_common">
                <param name="show_log" value="true"/>
            </section>
            <output name="hidden_output">
                <assert_contents>
                    <has_text_matching expression="snap.tl.umap"/>
                    <has_text_matching expression="n_comps = 2"/>
                    <has_text_matching expression="use_rep = 'X_spectral'"/>
                    <has_text_matching expression="key_added = 'umap'"/>
                    <has_text_matching expression="random_state = 0"/>
                </assert_contents>
            </output>
            <output name="anndata_out" ftype="h5ad">
                <assert_contents>
                    <has_h5_keys keys="obsm/X_umap"/>
                </assert_contents>
            </output>
        </test>
        <test expect_num_outputs="2">
            <!-- pp.knn -->
            <conditional name="method">
                <param name="method" value="pp.knn"/>
                <param name="adata" location="https://zenodo.org/records/17512085/files/tl.umap.pbmc_500_chr21.h5ad"/>
                <param name="n_neighbors" value="50"/>
                <param name="use_rep" value="X_spectral"/>
                <param name="algorithm" value="kdtree"/>
                <param name="random_state" value="0"/>
            </conditional>
            <section name="advanced_common">
                <param name="show_log" value="true"/>
            </section>
            <output name="hidden_output">
                <assert_contents>
                    <has_text_matching expression="snap.pp.knn"/>
                    <has_text_matching expression="n_neighbors = 50"/>
                    <has_text_matching expression="use_rep = 'X_spectral'"/>
                    <has_text_matching expression="method = 'kdtree'"/>
                    <has_text_matching expression="inplace = True"/>
                    <has_text_matching expression="random_state = 0"/>
                </assert_contents>
            </output>
            <output name="anndata_out" ftype="h5ad">
                <assert_contents>
                    <has_h5_keys keys="obsp/distances"/>
                </assert_contents>
            </output>
        </test>
        <test expect_num_outputs="2">
            <!-- tl.leiden -->
            <conditional name="method">
                <param name="method" value="tl.leiden"/>
                <param name="adata" location="https://zenodo.org/records/17512085/files/pp.knn.pbmc_500_chr21.h5ad"/>
                <param name="resolution" value="2"/>
                <param name="objective_function" value="modularity"/>
                <param name="min_cluster_size" value="3"/>
                <param name="n_iterations" value="-1"/>
                <param name="random_state" value="0"/>
                <param name="key_added" value="leiden"/>
                <param name="weighted" value="False"/>
            </conditional>
            <section name="advanced_common">
                <param name="show_log" value="true"/>
            </section>
            <output name="hidden_output">
                <assert_contents>
                    <has_text_matching expression="snap.tl.leiden"/>
                    <has_text_matching expression="resolution = 2"/>
                    <has_text_matching expression="objective_function = 'modularity'"/>
                    <has_text_matching expression="min_cluster_size = 3"/>
                    <has_text_matching expression="n_iterations = -1"/>
                    <has_text_matching expression="random_state = 0"/>
                    <has_text_matching expression="key_added = 'leiden'"/>
                    <has_text_matching expression="weighted = False"/>
                </assert_contents>
            </output>
            <output name="anndata_out" ftype="h5ad">
                <assert_contents>
                    <has_h5_keys keys="obs/leiden"/>
                </assert_contents>
            </output>
        </test>
        <test expect_num_outputs="2">
            <!-- tl.leiden -->
            <conditional name="method">
                <param name="method" value="tl.leiden"/>
                <param name="adata" location="https://zenodo.org/records/17512085/files/pp.knn.pbmc_500_chr21.h5ad"/>
                <param name="resolution" value="2"/>
                <param name="objective_function" value="RBConfiguration"/>
                <param name="min_cluster_size" value="3"/>
                <param name="n_iterations" value="-1"/>
                <param name="random_state" value="0"/>
                <param name="key_added" value="leiden"/>
                <param name="weighted" value="False"/>
            </conditional>
            <section name="advanced_common">
                <param name="show_log" value="true"/>
            </section>
            <output name="hidden_output">
                <assert_contents>
                    <has_text_matching expression="snap.tl.leiden"/>
                    <has_text_matching expression="resolution = 2"/>
                    <has_text_matching expression="objective_function = 'RBConfiguration'"/>
                    <has_text_matching expression="min_cluster_size = 3"/>
                    <has_text_matching expression="n_iterations = -1"/>
                    <has_text_matching expression="random_state = 0"/>
                    <has_text_matching expression="key_added = 'leiden'"/>
                    <has_text_matching expression="weighted = False"/>
                </assert_contents>
            </output>
            <output name="anndata_out" ftype="h5ad">
                <assert_contents>
                    <has_h5_keys keys="obs/leiden"/>
                </assert_contents>
            </output>
        </test>
        <test expect_num_outputs="2">
            <!-- tl.kmeans -->
            <conditional name="method">
                <param name="method" value="tl.kmeans"/>
                <param name="adata" location="https://zenodo.org/records/17512085/files/tl.spectral.pbmc_500_chr21.h5ad"/>
                <param name="n_iterations" value="-1"/>
                <param name="random_state" value="0"/>
                <param name="use_rep" value="X_spectral"/>
                <param name="key_added" value="kmeans"/>
            </conditional>
            <section name="advanced_common">
                <param name="show_log" value="true"/>
            </section>
            <output name="hidden_output">
                <assert_contents>
                    <has_text_matching expression="snap.tl.kmeans"/>
                    <has_text_matching expression="n_iterations = -1"/>
                    <has_text_matching expression="random_state = 0"/>
                    <has_text_matching expression="use_rep = 'X_spectral'"/>
                    <has_text_matching expression="key_added = 'kmeans'"/>
                </assert_contents>
            </output>
            <output name="anndata_out" ftype="h5ad">
                <assert_contents>
                    <has_h5_keys keys="obs/kmeans"/>
                </assert_contents>
            </output>
        </test>
        <test expect_num_outputs="2">
            <!-- tl.dbscan -->
            <conditional name="method">
                <param name="method" value="tl.dbscan"/>
                <param name="adata" location="https://zenodo.org/records/17512085/files/tl.spectral.pbmc_500_chr21.h5ad"/>
                <param name="eps" value="0.5"/>
                <param name="min_samples" value="3"/>
                <param name="leaf_size" value="5"/>
                <param name="use_rep" value="X_spectral"/>
                <param name="key_added" value="dbscan"/>
            </conditional>
            <section name="advanced_common">
                <param name="show_log" value="true"/>
            </section>
            <output name="hidden_output">
                <assert_contents>
                    <has_text_matching expression="snap.tl.dbscan"/>
                    <has_text_matching expression="eps = 0.5"/>
                    <has_text_matching expression="min_samples = 3"/>
                    <has_text_matching expression="leaf_size = 5"/>
                    <has_text_matching expression="use_rep = 'X_spectral'"/>
                    <has_text_matching expression="key_added = 'dbscan'"/>
                </assert_contents>
            </output>
            <output name="anndata_out" ftype="h5ad">
                <assert_contents>
                    <has_h5_keys keys="obs/dbscan"/>
                </assert_contents>
            </output>
        </test>
        <test expect_num_outputs="2">
            <!-- tl.hdbscan -->
            <conditional name="method">
                <param name="method" value="tl.hdbscan"/>
                <param name="adata" location="https://zenodo.org/records/17512085/files/tl.spectral.pbmc_500_chr21.h5ad"/>
                <param name="min_cluster_size" value="3"/>
                <param name="min_samples" value="3"/>
                <param name="cluster_selection_method" value="eom"/>
                <param name="random_state" value="0"/>
                <param name="use_rep" value="X_spectral"/>
                <param name="key_added" value="hdbscan"/>
            </conditional>
            <section name="advanced_common">
                <param name="show_log" value="true"/>
            </section>
            <output name="hidden_output">
                <assert_contents>
                    <has_text_matching expression="snap.tl.hdbscan"/>
                    <has_text_matching expression="min_cluster_size = 3"/>
                    <has_text_matching expression="min_samples = 3"/>
                    <has_text_matching expression="cluster_selection_method = 'eom'"/>
                    <has_text_matching expression="random_state = 0"/>
                    <has_text_matching expression="use_rep = 'X_spectral'"/>
                    <has_text_matching expression="key_added = 'hdbscan'"/>
                </assert_contents>
            </output>
            <output name="anndata_out" ftype="h5ad">
                <assert_contents>
                    <has_h5_keys keys="obs/hdbscan"/>
                </assert_contents>
            </output>
        </test>
        <!-- <test expect_num_outputs="2">
            tl.aggregate_X
            <conditional name="method">
                <param name="method" value="tl.aggregate_X"/>
                <param name="adata" location="https://zenodo.org/records/17512085/files/tl.spectral.pbmc_500_chr21.h5ad"/>
                <param name="normalize" value="RPKM"/>
            </conditional>
            <section name="advanced_common">
                <param name="show_log" value="true"/>
            </section>
            <output name="hidden_output">
                <assert_contents>
                    <has_text_matching expression="snap.tl.aggregate_X"/>
                    <has_text_matching expression="normalize = 'RPKM'"/>
                </assert_contents>
            </output>
            <output name="anndata_out" ftype="h5ad">
                <assert_contents>
                    <has_h5_keys keys="obs/n_fragment"/>
                </assert_contents>
            </output>
        </test> -->
        <test expect_num_outputs="2">
            <!-- tl.aggregate_cells -->
            <conditional name="method">
                <param name="method" value="tl.aggregate_cells"/>
                <param name="adata" location="https://zenodo.org/records/17512085/files/tl.spectral.pbmc_500_chr21.h5ad"/>
                <param name="use_rep" value="X_spectral"/>
                <param name="target_num_cells" value="5"/>
                <param name="min_cluster_size" value="3"/>
                <param name="random_state" value="0"/>
                <param name="key_added" value="pseudo_cell"/>
            </conditional>
            <section name="advanced_common">
                <param name="show_log" value="true"/>
            </section>
            <output name="hidden_output">
                <assert_contents>
                    <has_text_matching expression="snap.tl.aggregate_cells"/>
                    <has_text_matching expression="use_rep = 'X_spectral'"/>
                    <has_text_matching expression="target_num_cells = 5"/>
                    <has_text_matching expression="min_cluster_size = 3"/>
                    <has_text_matching expression="random_state = 0"/>
                    <has_text_matching expression="key_added = 'pseudo_cell'"/>
                </assert_contents>
            </output>
            <output name="anndata_out" ftype="h5ad">
                <assert_contents>
                    <has_h5_keys keys="obs/pseudo_cell"/>
                </assert_contents>
            </output>
        </test>
    </tests>
    <help><![CDATA[
Perform dimension reduction using Laplacian Eigenmap, using `tl.spectral`
=========================================================================

Perform dimension reduction using Laplacian Eigenmaps.

Convert the cell-by-feature count matrix into lower dimensional representations using the spectrum of the normalized graph Laplacian defined by pairwise similarity between cells. This function utilizes the matrix-free spectral embedding algorithm to compute the embedding when `distance_metric` is “cosine”, which scales linearly with the number of cells. For other types of similarity metrics, the time and space complexity scale quadratically with the number of cells.

More details on the `SnapATAC2 documentation
<https://scverse.org/SnapATAC2/api/_autosummary/snapatac2.tl.spectral.html>`__

Compute Laplacian Eigenmaps simultaneously on multiple modalities, with linear space and time complexity, using `tl.multi_spectral`
===================================================================================================================================

This is similar to `spectral`, but it can work on multiple modalities.

More details on the `SnapATAC2 documentation
<https://scverse.org/SnapATAC2/api/_autosummary/snapatac2.tl.multi_spectral.html>`__

Compute Umap, using `tl.umap`
=============================

Compute Umap

More details on the `SnapATAC2 documentation
<https://scverse.org/SnapATAC2/api/_autosummary/snapatac2.tl.umap.html>`__

Compute a neighborhood graph of observations, using `pp.knn`
============================================================

Compute a neighborhood graph of observations.

Computes a neighborhood graph of observations stored in adata using the method specified by method. The distance metric used is Euclidean.

More details on the `SnapATAC2 documentation
<https://scverse.org/SnapATAC2/api/_autosummary/snapatac2.pp.knn.html>`__

Cluster cells into subgroups, using `tl.leiden`
===============================================

Cluster cells into subgroups.

Cluster cells using the Leiden algorithm, an improved version of the Louvain algorithm. It has been proposed for single-cell analysis by. This requires having ran `knn`.

More details on the `SnapATAC2 documentation
<https://scverse.org/SnapATAC2/api/_autosummary/snapatac2.tl.leiden.html>`__

Cluster cells into subgroups using the K-means algorithm, using `tl.kmeans`
===========================================================================

Cluster cells into subgroups using the K-means algorithm, a classical algorithm in data mining.

More details on the `SnapATAC2 documentation
<https://scverse.org/SnapATAC2/api/_autosummary/snapatac2.tl.kmeans.html>`__

Cluster cells into subgroups using the DBSCAN algorithm, using `tl.dbscan`
==========================================================================

Cluster cells into subgroups using the DBSCAN algorithm.

More details on the `SnapATAC2 documentation
<https://scverse.org/SnapATAC2/api/_autosummary/snapatac2.tl.dbscan.html>`__

Cluster cells into subgroups using the HDBSCAN algorithm, using `tl.hdbscan`
============================================================================

Cluster cells into subgroups using the HDBSCAN algorithm.

More details on the `SnapATAC2 documentation
<https://scverse.org/SnapATAC2/api/_autosummary/snapatac2.tl.hdbscan.html>`__

.. Aggregate values in adata.X in a row-wise fashion, using `tl.aggregate_X`
.. =========================================================================

.. Aggregate values in adata.X in a row-wise fashion.

.. Aggregate values in adata.X in a row-wise fashion. This is used to compute RPKM or RPM values stratified by user-provided groupings.

.. More details on the `SnapATAC2 documentation
.. <https://scverse.org/SnapATAC2/api/_autosummary/snapatac2.tl.aggregate_X.html>`__

Aggregate cells into pseudo-cells, using `tl.aggregate_cells`
=============================================================

Aggregate cells into pseudo-cells.

Aggregate cells into pseudo-cells by iterative clustering.

More details on the `SnapATAC2 documentation
<https://scverse.org/SnapATAC2/api/_autosummary/snapatac2.tl.aggregate_cells.html>`__
    ]]></help>
    <expand macro="citations"/>
</tool>