autogenes.Interface.deconvolve

Interface.deconvolve(bulk, key = None, model='nusvr')

Performs bulk deconvolution

Deconvolves bulk data using a gene selection. The selection can be specified through a key or the current selection is used.

If the optimizer has been run, but nothing has been selected yet, an automatic selection occurs (equivalent to ag.select())

Parameters

• bulk (np.ndarray, pd.Series, pd.DataFrame, AnnData) – If multi-dimensional, then each row corresponds to a sample. If it has gene annotations (e.g. var_names for AnnData or df.columns for DataFrame), the method will respond intelligently (reorder if necessary, use only those genes from the selection that are available in the bulk data)

• key (str, optional (default: None)) – Name of the var column that specifies a gene selection. If None, then the current selection is used (or is automatically chosen)

• model (nusvr, nnls, linear, optional (default: nusvr)) – Choose a regression model. Available options: NuSVR, non-negative least squares and linear model.

Returns

An array of the form [[float, …],…] containing the model coefficients for each target (bulk sample)