Parameters Reference¶

This page documents all parameters for scPPIN-py methods with guidance on when and how to use them.

detect_module() Parameters¶

The detect_module() method has several parameters that control module detection:

detect_module() Parameters¶
Parameter	Default	Description
`fdr`	0.01	False discovery rate threshold. Lower values are more stringent and produce smaller modules. Increase if module is too small, decrease if too large.
`edge_weight_attr`	None	Edge attribute name containing weights (e.g., ‘weight’, ‘confidence’). If None, uses uniform edge costs matching the R implementation. Set this when you have confidence scores for interactions.
`c0`	0.01	Minimum edge cost to prevent numerical instability from zero-cost edges. Usually doesn’t need to be changed.
`normalization`	‘minmax’	Normalization method for edge weights: ‘minmax’ (scale to [0,1]), ‘log1p’ (log transform), ‘power’ (raise to power 6), or None (use weights directly). Only used when `edge_weight_attr` is provided.
`simplify`	True	Remove self-loops and parallel edges from network. Usually keep as True unless you need to preserve all edges.
`validate`	True	Validate network structure (check for empty network, disconnected components, etc.). Keep as True unless you’re certain your network is valid.
`use_max_prize_root`	False	If True, use the node with highest prize as the PCST root. This makes results more deterministic but may miss optimal solutions. Use False for standard behavior.

When to Adjust Parameters¶

Module too small? * Increase fdr (e.g., 0.05) to include more genes * Set edge_weight_attr if you have confidence scores * Try use_max_prize_root=True for different root selection

Module too large? * Decrease fdr (e.g., 0.001) for stricter filtering * Use edge weights to prioritize high-confidence interactions

Have edge confidence scores? * Set edge_weight_attr='weight' (or your attribute name) * Choose appropriate normalization method:

'minmax': Default, scales weights to [0, 1]

'log1p': Good for skewed distributions

'power': Emphasizes strong edges (raises to power 6)

None: Use weights directly (must already be in [0, 1])

load_network() Parameters¶

load_network() Parameters¶
Parameter	Default	Description
`source`		Network source: file path (str), list of edge tuples, pandas DataFrame, or igraph Graph object
`weight_column`	None	Column name in CSV/DataFrame to use as edge weights. If provided, loads weights from that column and sets as ‘weight’ attribute on edges.
`fmt`	‘auto’	File format hint: ‘auto’ (detect from extension), ‘csv’, ‘graphml’, or ‘gml’. Usually ‘auto’ works fine.

Examples:

# Load from CSV with edge weights
model.load_network('edges.csv', weight_column='confidence')

# Load from list
edges = [('GENE1', 'GENE2'), ('GENE2', 'GENE3')]
model.load_network(edges)

# Load GraphML file
model.load_network('network.graphml', fmt='graphml')

set_node_weights() Parameters¶

set_node_weights() Parameters¶
Parameter	Default	Description
`weights`		Dictionary mapping gene names to p-values, or AnnData object with rank_genes_groups results
`groupby`	None	Key in adata.obs for grouping labels (required when weights is AnnData). Example: ‘leiden’, ‘louvain’
`group`	None	Specific group/cluster to extract p-values for (required when weights is AnnData). Example: ‘0’, ‘1’, ‘cluster_A’

Examples:

# From dictionary
pvalues = {'TP53': 0.0001, 'MDM2': 0.001}
model.set_node_weights(pvalues)

# From AnnData (requires scanpy)
sc.tl.rank_genes_groups(adata, 'leiden', method='wilcoxon')
model.set_node_weights(adata, groupby='leiden', group='0')

set_edge_weights() Parameters¶

set_edge_weights() Parameters¶
Parameter	Default	Description
`weights`		Dictionary mapping (u, v) tuples to weight values. Only edges present in the network are set (automatically filtered).
`attr_name`	‘weight’	Edge attribute name to store weights. Use this name in `detect_module(edge_weight_attr='weight')`.

Example:

weights = {
    ('TP53', 'MDM2'): 0.9,
    ('TP53', 'CDKN1A'): 0.8,
}
model.set_edge_weights(weights=weights, attr_name='weight')
module = model.detect_module(fdr=0.01, edge_weight_attr='weight')

plot_module() Parameters¶

plot_module() Parameters¶
Parameter	Default	Description
`fdr`	0.01	FDR threshold for visualization (should match the FDR used in `detect_module()`)
`**kwargs`		Additional plotting arguments passed to matplotlib