Tutorial

In order to use decibel you need a basic knowledge of how scanpy and annData objects work.

Computing transcriptional noise

In order to work with decibel you need to first run scanpy and load a dataset in an annData object.

adata = sc.read('path_to_file/filename.h5ad')

Then process data (normalization, log-transformation, QC filtering on cells and genes).

sc.pp.normalize_total(adata, target_sum=1e4)
sc.pp.log1p(adata)
sc.pp.filter_genes(adata, min_cells=3)
sc.pp.filter_cells(adata, min_genes=100)

Run PCA, feature selection (triku), batch effect correction (harmony) and dimensionality reduction (UMAP).

sc.pp.pca(adata)
sc.pp.neighbors(adata)
tk.tl.triku(adata)
sc.pp.pca(adata)
sce.pp.harmony_integrate(adata, 'batch')
sc.pp.neighbors(adata, use_rep='X_pca_harmony')
sc.tl.umap(adata)

Compute transcriptional noise as in Enge et al, (2017).

dcb.enge_transcriptional_noise(adata, 'batch')