popv.algorithms.XGboost

popv.algorithms.XGboost#

class popv.algorithms.XGboost(batch_key='_batch_annotation', labels_key='_labels_annotation', layer_key=None, result_key='popv_xgboost_prediction', classifier_dict={})[source]#

Class to compute XGboost classifier.

Parameters:

  • batch_key (str | None (default: '_batch_annotation')) – Key in obs field of adata for batch information.

  • labels_key (str | None (default: '_labels_annotation')) – Key in obs field of adata for cell-type information.

  • layer_key (str | None (default: None)) – Key in layers field of adata used for classification. By default uses ‘X’ (log1p10K).

  • result_key (str | None (default: 'popv_xgboost_prediction')) – Key in obs in which celltype annotation results are stored.

  • classifier_dict (str | None (default: {})) – Dictionary to supply non-default values for XGboost classifier. Options at xgboost.train(). Default is {‘tree_method’: ‘hist’, ‘device’: ‘cuda’ if settings.cuml else ‘cpu’, ‘objective’: ‘multi:softprob’}.

Methods table#

compute_integration(adata)

Compute integration of adata inplace.

compute_umap(adata)

Compute UMAP embedding of adata inplace.

predict(adata)

Predict celltypes using XGboost.

Methods#

abstractmethod XGboost.compute_integration(adata)[source]#

Compute integration of adata inplace.

abstractmethod XGboost.compute_umap(adata)[source]#

Compute UMAP embedding of adata inplace.

XGboost.predict(adata)[source]#

Predict celltypes using XGboost.

Parameters:

adata – Anndata object. Results are stored in adata.obs[self.result_key].

Contents