log1p#

bullkpy.pp.log1p(adata, *, layer='cpm', out_layer='log1p_cpm', inplace_X=False)[source]#

log1p-transform a layer (default: CPM) and store as a new layer.

Works with dense or sparse matrices.

Log-transform expression values using log(1 + x).

log1p applies a natural log–transformation to an expression matrix, which reduces the influence of highly expressed genes and makes expression values more suitable for visualization, clustering, and batch correction.

This function is typically applied after CPM normalization.

What it does#

•	Applies log(1 + x) to expression values
•	Works with both dense and sparse matrices
•	Writes results to a new layer by default
•	Optionally overwrites adata.X

Parameters#

adata
AnnData object containing normalized expression data.

layer
Input layer to transform.
If None, uses adata.X.
Default: “cpm”.

out_layer
Name of the layer where log-transformed values are stored.
Default: “log1p_cpm”.

inplace_X
If True, also writes the transformed matrix to adata.X. Default: False.

Recommended workflow#

A standard bulk RNA-seq preprocessing pipeline:

bk.pp.set_raw_counts(adata)
bk.pp.normalize_cpm(adata)
bk.pp.log1p(adata)

After this, log-transformed values are available in:

adata.layers["log1p_cpm"]

Examples#

Default usage (CPM → log1p)#

bk.pp.log1p(adata)

Transform a custom layer#

bk.pp.log1p(
    adata,
    layer="cpm",
    out_layer="log_expr"
)

Overwrite adata.X#

bk.pp.log1p(
    adata,
    inplace_X=True
)

Notes#

•	log1p is preferred over log(x) because it is well-defined for zero counts.
•	This transformation assumes input data are non-negative.
•	For batch correction and visualization (e.g. PCA, violin plots), log1p

expression is recommended.