scHumanNet API Reference

May 29, 2025 ยท View on GitHub

Python API

Core Functions

sort_add_lls(celltype_specific_networks, reference_network=None)

Sort SCINET output edges alphabetically and add LLS weight from HumanNetv3.

Parameters:

  • celltype_specific_networks (Dict[str, pd.DataFrame]): Dictionary of cell-type specific networks
  • reference_network (nx.Graph, optional): Reference network (HumanNetv3)

Returns:

  • Dict[str, pd.DataFrame]: Dictionary of dataframes with gene interactions and weights

get_centrality(method, net_list, exclude_ribosomal=True)

Calculate centrality values for nodes in scHumanNet networks.

Parameters:

  • method (str): Centrality measure ('degree', 'betweenness', 'closeness', 'eigenvector')
  • net_list (Dict[str, pd.DataFrame]): Output of sort_add_lls()
  • exclude_ribosomal (bool): Whether to exclude ribosomal genes

Returns:

  • Dict[str, Dict[str, float]]: Dictionary of centrality values for each cell type

combine_perc_rank(perc_rank_list, reference_genes=None)

Combine percentile rank centrality values across cell types.

Parameters:

  • perc_rank_list (Dict[str, Dict[str, float]]): Output of get_centrality()
  • reference_genes (List[str], optional): List of reference genes to include

Returns:

  • pd.DataFrame: DataFrame with genes as rows and cell types as columns

find_all_hub(net_list, centrality='degree', q_method='BH', threshold=0.05)

Find statistically significant hub genes in each network.

Parameters:

  • net_list (Dict[str, pd.DataFrame]): Output of sort_add_lls()
  • centrality (str): Centrality method
  • q_method (str): Multiple testing correction method
  • threshold (float): Significance threshold

Returns:

  • pd.DataFrame: DataFrame with significant hub genes

find_diff_hub(rank_df, meta, celltypes_col, condition_col, control_value, net_list, **kwargs)

Find statistically significant differential hub genes.

Parameters:

  • rank_df (pd.DataFrame): Output of combine_perc_rank()
  • meta (pd.DataFrame): Metadata DataFrame
  • celltypes_col (str): Column name for cell types
  • condition_col (str): Column name for conditions
  • control_value (str): Control condition value
  • net_list (Dict[str, pd.DataFrame]): Network list from sort_add_lls()

Returns:

  • pd.DataFrame: DataFrame with statistical significance results

Utility Functions

load_humannet_v3(data_path=None)

Load HumanNetv3 reference network.

connectivity(gene_list, network, method='mean')

Calculate connectivity measure for a gene list in a network.

network_stats(network)

Calculate basic network statistics.

save_networks(network_list, output_path, format='csv')

Save networks to files.

load_networks(input_path, format='csv')

Load networks from files.

R API

Core Functions

SortAddLLS(Celltype.specific.networks, reference.network)

Sort SCINET output edges alphabetically and add LLS weight derived from HumanNetv3.

Parameters:

  • Celltype.specific.networks: SCINET output from function run.SCINET.clusters()
  • reference.network: reference network input, default is HumanNetv3

Returns:

  • List of dataframe edgelist, with gene interaction and weights from SCINET and HumanNetv3

GetCentrality(method, net.list)

Get centrality values for each nodes of scHumanNet list.

Parameters:

  • method: Centrality measure ('degree', 'betweenness', 'closeness', 'eigenvector')
  • net.list: Output of SortAddLLS()

Returns:

  • List of named vector, each value corresponding to node's centrality value

CombinePercRank(perc.rank.list)

Calculate percentile rank of each centrality measure.

Parameters:

  • perc.rank.list: Output of GetCentrality()

Returns:

  • Dataframe of celltypes and their centrality values

FindAllHub(net.list, centrality='degree', q.method='BH', threshold=0.05)

Find statistically significant hub genes in each scHumanNets.

Parameters:

  • net.list: output of SortAddLLS
  • centrality: centrality method
  • q.method: multiple testing correction method
  • threshold: significance threshold

Returns:

  • Dataframe with Percentile Rank Centrality, gene, pvalue, qvalue and celltype

FindDiffHub(rank.df.final, meta, celltypes, condition, control, net.list, **kwargs)

Calculate statistical significance of diffPR values.

Parameters:

  • rank.df.final: Output from CombinePercRank
  • meta: metadata data.frame
  • celltypes: column name for celltypes annotation
  • condition: column name for disease and control annotation
  • control: character string specifying control value
  • net.list: Output of SortAddLLS

Returns:

  • Dataframe with percentile rank of centrality, differences, and statistical significance

Helper Functions

TopHub(rank.df.final, top.n)

Get top n genes in terms of centrality for each scHumanNet.

DiffPR(rank.df.final, celltypes, condition, control, meta)

Get difference of normalized centrality values.

Command Line Interface

The Python package includes a command-line interface:

# Find hub genes
schumannet find-hubs --networks /path/to/networks --output results.csv

# Differential analysis  
schumannet diff-analysis --networks /path/to/networks --metadata meta.csv \
  --celltype-col celltype --condition-col condition --control Control \
  --output diff_results.csv

# Calculate network statistics
schumannet network-stats --networks /path/to/networks --output stats.csv

# Convert R data to Python format
schumannet convert --input data.rda --output data.pkl

Data Formats

Network Format

Networks should be provided as DataFrames/data.frames with columns:

  • gene1: First gene in interaction
  • gene2: Second gene in interaction
  • LLS: LLS weight from HumanNetv3
  • scinet_weight: Weight from SCINET (optional)

Metadata Format

Metadata should contain:

  • Cell identifiers
  • Cell type annotations
  • Condition/group annotations
  • Additional covariates (optional)