Mapping Functions

class transbrain.trans.SpeciesTrans(atlas_type='bn')[source]

Bases: object

Transfer phenotypes between species using graph embeddings.

Parameters:

atlas_type ({'bn', 'dk', 'aal'}, optional) –

The type of human atlas used for initialization.

  • ’bn’ : Brainnetome Atlas

  • ’dk’ : Desikan-Killiany Atlas

  • ’aal’ : Automated Anatomical Labeling (AAL) Atlas

Default is ‘bn’.

atlas_type

The selected atlas type.

Type:

str

regions

Dictionary containing human and mouse brain regions (cortex, subcortex, all).

Type:

dict

embeddings

Loaded graph embeddings used for phenotype translation.

Type:

np.ndarray

_load_embeddings()[source]

Load graph embeddings for phenotype translation.

The function loads precomputed embeddings from a binary file (pickle format) based on the selected atlas type. These embeddings are used to map phenotypes between species.

Returns:

A NumPy array containing the loaded embeddings corresponding to the specified atlas type.

Return type:

np.ndarray

_dual_mapping(pheno_data, source_matrix, target_matrix, normalize=False)[source]

Map phenotype data from source to target space using dual regression.

Parameters:

  • pheno_data (np.ndarray) – An array of phenotype values (regions,) in the source species.

  • source_matrix (np.ndarray) – The embedding matrix for the source species.

  • target_matrix (np.ndarray) – The embedding matrix for the target species.

  • normalize (bool, optional) – Whether to normalize the phenotype values before regression. Default is False.

Returns:

An array of predicted phenotype values in the target species.

Return type:

np.ndarray

_translate(phenotype, direction, region_type='cortex', normalize=True, restore=False)[source]

Unified translation method for both directions.

Parameters:

  • phenotype (pd.DataFrame) – A DataFrame where rows are brain regions and columns are phenotype types.

  • direction ({'human_to_mouse', 'mouse_to_human'}) – The translation direction.

  • region_type ({'cortex', 'subcortex', 'all'}, optional) – The region subset to use for translation. Default is ‘cortex’.

  • normalize (bool, optional) – Whether to normalize phenotype values before translation. Default is True.

  • restore (bool, optional) – Whether to inverse-transform values back to original scale. Only used if normalize is True.

Returns:

Translated phenotype values in the target species, indexed by brain region name.

Return type:

pd.DataFrame

mouse_to_human(phenotype, region_type='cortex', normalize=True, restore=False)[source]

Translate mouse phenotype to human.

Parameters:

  • phenotype (pd.DataFrame) – Mouse phenotype DataFrame (regions × phenotypes).

  • region_type ({'cortex', 'subcortex', 'all'}, optional) – The brain region type to translate. Default is ‘cortex’.

  • normalize (bool, optional) – Whether to normalize data before translation. Default is True.

  • restore (bool, optional) – Whether to restore values back to original scale after translation. Only used if normalize is True.

Returns:

Translated human phenotype DataFrame (regions × phenotypes).

Return type:

pd.DataFrame

human_to_mouse(phenotype, region_type='cortex', normalize=True, restore=False)[source]

Translate human phenotype to mouse.

Parameters:

  • phenotype (pd.DataFrame) – Human phenotype DataFrame (regions × phenotypes).

  • region_type ({'cortex', 'subcortex', 'all'}, optional) – The brain region type to translate. Default is ‘cortex’.

  • normalize (bool, optional) – Whether to normalize data before translation. Default is True.

  • restore (bool, optional) – Whether to restore values back to original scale after translation. Only used if normalize is True.

Returns:

Translated mouse phenotype DataFrame (regions × phenotypes).

Return type:

pd.DataFrame