Tutorial

Tutorial#

Step 1: Clonal nearest neighbours graph construction#

Firstly, we will create AnnData-object with clones.

import scanpy as sc
import clone2vec as c2v

clones = c2v.pp.clones_adata(
    adata,
    obs_name="clone", # Column with clonal labels
    min_size=2,       # Minimal clone size
    na_value="NA",    # Value for non-labelled cells
)

Minimal clone size parameter is used to exclude small clones from embedding construction.

Step 2: clone2vec#

Now, we have to train our neural network to predict clonal labels of nearest neighbours for each clonally labelled cell.

c2v.tl.clonal_nn(
    adata,
    clones,
    use_rep="X_pca", # Which dimred to use for graph construction
)

c2v.tl.clone2vec(clones)

After execution of this function we have AnnData-object clones with clonal vector representation stored in clones.obsm["clone2vec"]. Now we can work with it like with regular scRNA-Seq dataset.

Step 3: clone2vec analysis#

sc.pp.neighbors(clones, use_rep="clone2vec")
sc.tl.umap(clones)
sc.tl.leiden(clones)

And after perform all other additional steps of analysis.

For a more detailed walkthrough see the Examples section.

Contents