### Install Decoupler (Development Version)

Source: https://github.com/scverse/decoupler/blob/main/README.md

Installs the latest development version of decoupler directly from its GitHub repository. This is useful for testing the newest features or contributing to the project.

```bash
pip install git+https://github.com/scverse/decoupler.git@main
```

--------------------------------

### Install Decoupler (PyPI - Stable)

Source: https://github.com/scverse/decoupler/blob/main/README.md

Installs the latest stable release of the decoupler package from PyPI with minimal dependencies. This is the recommended method for most users.

```bash
pip install decoupler
```

--------------------------------

### Install Decoupler (PyPI - Full Release)

Source: https://github.com/scverse/decoupler/blob/main/README.md

Installs the latest stable full release of the decoupler package from PyPI, including extra dependencies. Use this if you need additional features or integrations.

```bash
pip install decoupler[full]
```

--------------------------------

### Install Decoupler (Conda-Forge)

Source: https://github.com/scverse/decoupler/blob/main/README.md

Installs the latest stable version of decoupler from the conda-forge channel using mamba or conda. Note the '-py' suffix in the package name. This method is suitable for users within the Conda ecosystem.

```bash
mamba create -n=dcp conda-forge::decoupler-py
```

--------------------------------

### Get Results from obsm

Source: https://context7.com/scverse/decoupler/llms.txt

Explains how to extract enrichment results stored in the 'obsm' slot of an AnnData object as a new AnnData object, suitable for further analysis and visualization.

```python
import decoupler as dc
import scanpy as sc
import matplotlib.pyplot as plt

# Run enrichment analysis
adata, net = dc.ds.toy(nobs=100, nvar=30)
dc.mt.ulm(adata, net, tmin=3)

# Extract scores as new AnnData
scores_adata = dc.pp.get_obsm(adata, key='score_ulm')
print(scores_adata.shape)

# Now can use scanpy functions
sc.pp.scale(scores_adata)
sc.tl.pca(scores_adata)
sc.pp.neighbors(scores_adata)
sc.tl.umap(scores_adata)

# Visualize
sc.pl.umap(scores_adata, color=scores_adata.var_names[0:3],
           cmap='RdBu_r', vcenter=0)

# Compare p-values
pvals_adata = dc.pp.get_obsm(adata, key='padj_ulm')
print(f"Mean significant sources per cell: {(pvals_adata.X < 0.05).sum(axis=1).mean():.2f}")
```

--------------------------------

### Generate Toy Dataset for decoupler

Source: https://context7.com/scverse/decoupler/llms.txt

Generates synthetic AnnData and network objects for testing decoupler functionalities. It supports basic toy data generation and can simulate pseudotime for more complex scenarios. Outputs include AnnData with expression data and a Pandas DataFrame representing the network.

```python
import decoupler as dc

# Basic toy dataset with 30 samples and 20 features
adata, net = dc.ds.toy(nobs=30, nvar=20, bval=2.0, seed=42)
print(adata.shape)  # (30, 20)
print(net.head())
#   source target  weight
# 0     T1    G01     1.0
# 1     T1    G02     1.0
# 2     T1    G03     0.7
# 3     T2    G04     1.0
# 4     T2    G06    -0.5

# Generate toy dataset with pseudotime simulation
adata_pt, net_pt = dc.ds.toy(nobs=50, nvar=25, pstime=True, verbose=True)
print(adata_pt.obs['pstime'].head())
# C01    0.00
# C02    0.02
# C03    0.04
# ...

```

--------------------------------

### Extract Data from AnnData

Source: https://context7.com/scverse/decoupler/llms.txt

Illustrates extracting expression matrices from AnnData objects, including options for different layers and raw data. It also shows how methods handle extraction internally.

```python
import decoupler as dc
import scanpy as sc
import scipy.sparse as sp

# Load data
adata = sc.datasets.pbmc3k_processed()

# Extract data from default layer (X)
mat, obs_names, var_names = dc.pp.extract(adata, layer=None,
                                          use_raw=False)
print(mat.shape)
print(f"Type: {type(mat)}")

# Extract from specific layer
adata.layers['counts'] = adata.raw.X.copy()
mat_counts, _, _ = dc.pp.extract(adata, layer='counts')

# Convert to dense if needed
if sp.issparse(mat_counts):
    mat_counts_dense = mat_counts.toarray()

# Use with methods
_, net = dc.ds.toy()
# Methods automatically handle extraction internally
dc.mt.ulm(adata, net, layer=None, use_raw=False, tmin=5)
```

--------------------------------

### Shuffle Network and Compute Correlation

Source: https://context7.com/scverse/decoupler/llms.txt

Demonstrates shuffling a network for null distribution and computing network correlation using decoupler's utility functions.

```python
import decoupler as dc

# Shuffle network for null distribution
shuffled_net = dc.pp.shuffle_net(net, target='target', seed=42)

# Compute network correlation
net_with_cor = dc.pp.net_corr(mat=adata.X, net=net, verbose=True)
print(net_with_cor.head())
```

--------------------------------

### Build Adjacency Matrix

Source: https://context7.com/scverse/decoupler/llms.txt

Shows how to build an adjacency matrix from expression data and a network using the adjmat function.

```python
import decoupler as dc

# Build adjacency matrix for methods
adj_matrix = dc.pp.adjmat(mat=adata.to_df(), net=net, verbose=True)
print(adj_matrix.shape)
```

--------------------------------

### Visualize Enrichment Results

Source: https://context7.com/scverse/decoupler/llms.txt

Demonstrates various plotting functions in decoupler for visualizing enrichment scores, including dotplots, barplots, volcano plots, and network visualizations.

```python
import decoupler as dc
import matplotlib.pyplot as plt

# Run analysis
adata, net = dc.ds.toy(nobs=50, nvar=25, seed=42)
adata.obs['condition'] = (['A'] * 25) + (['B'] * 25)
dc.mt.ulm(adata, net, tmin=3)

# Dotplot of activities
dc.pl.dotplot(adata, groupby='condition', var_names=net['source'].unique(),
              obsm_key='score_ulm', title='TF Activities')
plt.tight_layout()
plt.show()

# Barplot for specific sample
dc.pl.barplot(adata, obsm_key='score_ulm', obs_name='C01',
              top=10, title='Top 10 Activities')
plt.show()

# Volcano plot comparing groups
dc.tl.rankby_group(adata, groupby='condition', groups=['A'],
                   reference='B', obsm='score_ulm')
dc.pl.volcano(adata, key='rankby_group')
plt.show()

# Network visualization
dc.pl.network(net, n_sources=3, n_targets=10,
              source=['T1', 'T2', 'T3'])
plt.show()
```

--------------------------------

### Run Multiple Enrichment Methods (Decouple) in decoupler

Source: https://context7.com/scverse/decoupler/llms.txt

Executes multiple enrichment methods sequentially within the decoupler package and optionally computes consensus scores. The `decouple` function can run all available methods or a specified subset, storing results in adata.obsm. It requires an AnnData object and a network.

```python
import decoupler as dc

# Load data
adata, net = dc.ds.toy(nobs=30, nvar=25)

# Run all available methods
dc.mt.decouple(adata, net, methods='all', tmin=3, verbose=True)

# Check all results stored
print([k for k in adata.obsm.keys() if 'score_' in k])
# ['score_aucell', 'score_gsea', 'score_gsva', 'score_mlm',
#  'score_ulm', 'score_viper', 'score_waggr', 'score_zscore']

```

--------------------------------

### Translate Gene Symbols

Source: https://context7.com/scverse/decoupler/llms.txt

Shows how to translate gene symbols between different organisms using the `translate` function. It includes displaying available organisms and handling a sample DataFrame.

```python
import decoupler as dc
import pandas as pd

# Show available organisms
organisms = dc.op.show_organisms()
print(organisms['name'].unique()[:10])

# Create sample data with human genes
df = pd.DataFrame({
    'genesymbol': ['TP53', 'BRCA1', 'EGFR', 'MYC', 'KRAS'],
    'value': [1.5, 2.3, -0.8, 1.9, -1.2]
})

# Translate from human to mouse
df_mouse = dc.op.translate(df, columns='genesymbol',
                           target_organism='mouse', verbose=True)
print(df_mouse)
```

--------------------------------

### Run Decoupling Methods with Custom Parameters and Consensus

Source: https://context7.com/scverse/decoupler/llms.txt

Execute specific decoupling methods ('ulm', 'mlm', 'viper') with custom arguments or enable consensus scoring for aggregated results. This function takes an AnnData object and a network as input, allowing flexible analysis.

```python
import decoupler as dc

# Run specific methods with custom parameters
methods_to_run = ['ulm', 'mlm', 'viper']
args = {
    'ulm': {'tval': True},
    'mlm': {'tval': False},
    'viper': {'nes': True}
}
dc.mt.decouple(adata, net, methods=methods_to_run, args=args,
               tmin=5, verbose=True)

# Run with consensus scoring
adata2, net2 = dc.ds.toy(nobs=30, nvar=25)
dc.mt.decouple(adata2, net2, methods=['ulm', 'mlm', 'viper'],
               cons=True, tmin=3)
consensus = adata2.obsm['score_consensus']
print(consensus.head())
```

--------------------------------

### Access OmniPath Resources and Databases

Source: https://context7.com/scverse/decoupler/llms.txt

Query various biological resources through the OmniPath database, including PanglaoDB (cell markers), CollecTRI (TF-target interactions), and MSigDB Hallmark gene sets. This provides access to curated biological data for analysis.

```python
import decoupler as dc

# Show available resources
resources = dc.op.show_resources()
print(resources.head(10))

# Get specific resource (e.g., PanglaoDB cell markers)
panglaodb = dc.op.resource(name='PanglaoDB', organism='human',
                            license='academic', verbose=True)
print(panglaodb.head())

# Get CollecTRI network
collectri = dc.op.collectri(organism='human', verbose=True)
print(f"CollecTRI interactions: {len(collectri)}")

# Get MSigDB Hallmark gene sets
hallmark = dc.op.hallmark(organism='human', verbose=True)
print(hallmark['source'].unique()[:5])
```

--------------------------------

### Translate Network Between Organisms using Decoupler

Source: https://context7.com/scverse/decoupler/llms.txt

Translates a given network from one organism to another, specifying source and target columns. Requires the decoupler library. Handles organism translation for network data.

```python
import decoupler as dc

human_net = dc.op.dorothea(organism='human', levels=['A', 'B'])
mouse_net = dc.op.translate(human_net, columns=['source', 'target'],
                            target_organism='mouse', verbose=True)
```

--------------------------------

### Compute Consensus Score Across Multiple Methods

Source: https://context7.com/scverse/decoupler/llms.txt

Calculate consensus enrichment scores and p-values by combining results from multiple decoupling methods. This involves running decouple, then computing consensus using `dc.mt.consensus`, and visualizing the results.

```python
import decoupler as dc
import matplotlib.pyplot as plt

# Load data and run multiple methods
adata, net = dc.ds.toy(nobs=25, nvar=20)
dc.mt.decouple(adata, net, methods=['ulm', 'mlm', 'waggr'], tmin=3)

# Compute consensus manually
dc.mt.consensus(adata, verbose=True)

# Access consensus results
consensus_scores = adata.obsm['score_consensus']
consensus_pvals = adata.obsm['padj_consensus']

# Compare consensus with individual methods
fig, axes = plt.subplots(1, 3, figsize=(15, 4))
adata.obsm['score_ulm'].iloc[0].plot(kind='bar', ax=axes[0], title='ULM')
adata.obsm['score_mlm'].iloc[0].plot(kind='bar', ax=axes[1], title='MLM')
consensus_scores.iloc[0].plot(kind='bar', ax=axes[2], title='Consensus')
plt.tight_layout()
plt.show()
```

--------------------------------

### Filter AnnData by Expression

Source: https://context7.com/scverse/decoupler/llms.txt

Demonstrates filtering genes from an AnnData object based on expression levels and proportions, with options for in-place modification and grouping.

```python
import decoupler as dc
import scanpy as sc

# Load data
adata = sc.datasets.pbmc3k_processed()
print(f"Before filtering: {adata.shape}")

# Filter by expression level (at least 10 cells with min counts)
dc.pp.filter_by_expr(adata, group=None, min_count=10, min_total_count=15,
                     inplace=True)
print(f"After expression filter: {adata.shape}")

# Filter by proportion (expressed in at least 10% of cells)
adata2 = sc.datasets.pbmc3k_processed()
dc.pp.filter_by_prop(adata2, min_prop=0.1, min_smpls=5, inplace=True)
print(f"After proportion filter: {adata2.shape}")

# Use with grouping variable
adata3 = sc.datasets.pbmc3k_processed()
dc.pp.filter_by_expr(adata3, group='louvain', min_count=5,
                     min_total_count=10, inplace=True)
print(f"After group-aware filter: {adata3.shape}")
```

--------------------------------

### Preprocess Network Data with Pruning

Source: https://context7.com/scverse/decoupler/llms.txt

Filter and manipulate network data before enrichment analysis using functions like `dc.pp.prune`. This function removes sources (e.g., transcription factors) that have fewer than a specified minimum number of targets.

```python
import decoupler as dc
import numpy as np

# Load data
adata, net = dc.ds.toy(nobs=30, nvar=30)

# Prune network - remove sources with fewer than tmin targets
pruned_net = dc.pp.prune(features=adata.var_names, net=net,
                         tmin=3, verbose=True)
print(f"Original sources: {net['source'].nunique()}")
print(f"Pruned sources: {pruned_net['source'].nunique()}")

# Read gene sets from GMT file
# net = dc.pp.read_gmt("path/to/geneset.gmt")
```

--------------------------------

### Access PROGENy Pathway Activity Network

Source: https://context7.com/scverse/decoupler/llms.txt

Retrieve the PROGENy network, which maps transcription factors to 14 cancer-relevant pathways. This function allows specifying the organism and the number of top genes per pathway to consider.

```python
import decoupler as dc

# Get PROGENy network for human
progeny_net = dc.op.progeny(organism='human', top=500, verbose=True)
print(progeny_net['source'].unique())

print(progeny_net.head())

# Get top 100 genes per pathway
progeny_top100 = dc.op.progeny(organism='human', top=100)

# Mouse data
progeny_mouse = dc.op.progeny(organism='mouse', top=500)

# Apply to analyze pathway activities
adata, _ = dc.ds.toy(nobs=100, nvar=500)
dc.mt.mlm(adata, progeny_net, tmin=10, verbose=True)
pathway_scores = adata.obsm['score_mlm']
print(pathway_scores.head())
```

--------------------------------

### Access DoRothEA Gene Regulatory Network

Source: https://context7.com/scverse/decoupler/llms.txt

Retrieve the DoRothEA gene regulatory network, which contains transcription factor-target interactions. You can filter by confidence levels ('A', 'B', 'C') and customize weights for network analysis.

```python
import decoupler as dc
import scanpy as sc

# Get DoRothEA network with high-confidence interactions (A, B, C)
dorothea_net = dc.op.dorothea(organism='human', levels=['A', 'B', 'C'],
                               verbose=True)
print(dorothea_net.head())

print(f"Total interactions: {len(dorothea_net)}")
print(f"Unique TFs: {dorothea_net['source'].nunique()}")
print(f"Unique targets: {dorothea_net['target'].nunique()}")

# Use only highest confidence level
dorothea_A = dc.op.dorothea(levels='A', verbose=False)

# Custom weights by confidence level
custom_weights = {'A': 1, 'B': 3, 'C': 5}
dorothea_custom = dc.op.dorothea(levels=['A', 'B', 'C'],
                                  dict_weights=custom_weights)

# Apply to real data
adata = sc.datasets.pbmc3k_processed()
dc.mt.ulm(adata, dorothea_net, tmin=5, verbose=True)
```

--------------------------------

### Univariate Linear Model (ULM) in decoupler

Source: https://context7.com/scverse/decoupler/llms.txt

Fits univariate linear models to infer enrichment scores using correlation-based statistics within the decoupler package. It takes an AnnData object and a network as input and stores results (scores and adjusted p-values) in adata.obsm. Minimum target filtering can be applied.

```python
import decoupler as dc
import numpy as np

# Load data
adata, net = dc.ds.toy(nobs=50, nvar=30, seed=123)

# Run ULM with minimum target filtering
dc.mt.ulm(adata, net, tmin=3, verbose=True)

# Results stored in adata.obsm
print(adata.obsm.keys())
# dict_keys(['score_ulm', 'padj_ulm'])

# Access enrichment scores
scores = adata.obsm['score_ulm']
print(scores.head())
#         T1        T2        T3        T4        T5
# C01  8.234   -0.123    0.456   -1.234    2.345
# C02  7.891    0.234   -0.789   -2.456    1.987
# ...

# Access adjusted p-values
pvals = adata.obsm['padj_ulm']
print(f"Significant activities: {(pvals < 0.05).sum().sum()}")

```

--------------------------------

### Rank Gene Activities by Group using Decoupler

Source: https://context7.com/scverse/decoupler/llms.txt

Computes differential activity scores between groups of samples using decoupler's rankby_group function. It requires an AnnData object and a network. The function calculates scores and p-values for ranked genes, allowing visualization via barplots.

```python
import decoupler as dc

# Prepare data
adata, net = dc.ds.toy(nobs=60, nvar=30, seed=123)
adata.obs['celltype'] = (['TypeA'] * 20 + ['TypeB'] * 20 + ['TypeC'] * 20)

# Run enrichment
dc.mt.ulm(adata, net, tmin=3)

# Rank activities comparing TypeA vs rest
dc.tl.rankby_group(adata, groupby='celltype', groups=['TypeA'],
                   obsm='score_ulm', method='wilcoxon')

# Access results
rank_results = adata.uns['rankby_group']
print(rank_results['names'].head())  # Top ranked sources
print(rank_results['scores'].head())  # Test statistics
print(rank_results['pvals_adj'].head())  # Adjusted p-values

# Compare specific groups
dc.tl.rankby_group(adata, groupby='celltype', groups=['TypeA'],
                   reference='TypeB', obsm='score_ulm')

# Visualize
dc.pl.barplot(adata.uns['rankby_group'], top=5)
```

--------------------------------

### Over Representation Analysis (ORA) in decoupler

Source: https://context7.com/scverse/decoupler/llms.txt

Performs Fisher's exact test-based enrichment analysis for binary feature sets using decoupler. It takes an AnnData object and a network, allowing specification of columns for sources and targets. Outputs include log odds ratios and adjusted p-values stored in adata.obsm.

```python
import decoupler as dc
import numpy as np

# Create data with significant features
adata, net = dc.ds.toy(nobs=20, nvar=30)

# Mark top expressed genes as significant (e.g., from differential expression)
top_n = 10
pvals = np.random.uniform(0, 1, adata.n_vars)
pvals[:top_n] = 0.001  # Make first 10 genes significant
adata.var['pvals'] = pvals

# Run ORA using p-value threshold
dc.mt.ora(adata, net,
          source='pvals', target='pvals',
          min_n=3, verbose=True)

# Access ORA results
ora_scores = adata.obsm['score_ora']  # Log odds ratios
ora_pvals = adata.obsm['padj_ora']   # Fisher's exact test p-values

print(f"Enriched sources (padj < 0.05): {(ora_pvals < 0.05).sum(axis=1)}")

```

--------------------------------

### Multivariate Linear Model (MLM) in decoupler

Source: https://context7.com/scverse/decoupler/llms.txt

Fits multivariate linear models in decoupler to infer enrichment scores while accounting for simultaneous correlations between sources. It requires an AnnData object and a network, storing results in adata.obsm. This method can be compared against ULM results.

```python
import decoupler as dc

# Load data
adata, net = dc.ds.toy(nobs=40, nvar=25)

# Run MLM - accounts for dependencies between transcription factors
dc.mt.mlm(adata, net, tmin=5, verbose=True)

# Compare with ULM results
dc.mt.ulm(adata, net, tmin=5, verbose=False)

# Extract both results
mlm_scores = adata.obsm['score_mlm']
ulm_scores = adata.obsm['score_ulm']

# Check correlation between methods
import scipy.stats as sts
corr = sts.spearmanr(mlm_scores.values.flatten(),
                      ulm_scores.values.flatten())
print(f"Spearman correlation: {corr.correlation:.3f}, p={corr.pvalue:.2e}")

```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.