predict_interval_gradients

predict_interval_gradients#

DnaClient.predict_interval_gradients(gene_id, *, organism=Organism.HOMO_SAPIENS, requested_output, sequence_length=1048576, condition=None, cell_conditioning=None, cell_condition_name=None)[source]#

Generate input gradients of a gene’s predicted expression.

The input interval is centered on the gene’s median TSS.

Parameters:

  • gene_id (str) – Ensembl ID of the target gene. Version suffixes are ignored.

  • organism (Organism (default: <Organism.HOMO_SAPIENS: 9606>)) – Organism to use for the prediction.

  • requested_output (OutputType) – Output type to compute gradients for. One of RNA_SEQ, SC_GENE_EXPRESSION, or SC_GENE_EXPRESSION_DENOISED.

  • sequence_length (int (default: 1048576)) – Width of the input interval. Must be SEQUENCE_LENGTH_100KB or SEQUENCE_LENGTH_1MB (defaults to SEQUENCE_LENGTH_1MB).

  • condition (ModelCondition | str | None (default: None)) – The condition or composite curie to generate the prediction for.

  • cell_conditioning (AnnData | None (default: None)) – Optional expression conditioning input (AnnData).

  • cell_condition_name (str | None (default: None)) – Optional label for expression conditioning. When provided, the output track names are suffixed with this name. Must be provided when cell_conditioning is used.

Return type:

InputGradientsOutput

Returns:

Input gradients for the interval centered on the gene.