With the implementation of the new datastructure (#327), it becomes rather tricky to get information
from the airr rearrangment schema (e.g. what is the "c_call" of the primary "VJ" chain?).

Previously this was possible simply with

adata.obs["IR_VJ_1_c_call"]

With the new datastructure

adata[:, "c_call"].X

just yields an awkward array with a variable number of chains per cell. The information which
chain is VJ/VDJ and primary or secondary is hidden away in adata.obsm.

This motivates the implementation of long-proposed easy-access getter/setter functions. At the very least to

• retrieve AIRR rearrangement variables for a certain chain.

But possibly also with convenience functions, e.g.

• to retrieve the most abundant categories (previously discussed in Straightforward way for getting most abundant categories #51).

The latter is of less importance, but the interface needs to be designed jointly, therefore this is also a topic in this issue.

To get AIRR data, we need something like

ir.get(adata, "locus", "VJ_1") -> pd.Series

We need it regulary to get the top n of a category, e.g. v-gene or clonotype.

This can be achieved in a pandas onliner, if one knows how to do it...

# This is probably hacky, we might think about a better way, but we need the most abundant clonotypes

top_clonotypes = adata.obs.clonotype.value_counts()[:8].index.tolist() # A better way might be needed especailly to take normalization into account
top_vgenes = adata.obs.TRB_1_v_gene.value_counts()[:8].index.tolist()

It would be more user-friendly to have a convenience function to this,
for instance:

ir.tl.top_n(col="clonotype", n=10)

Use it for plotting:

sc.pl.umap(adata, color="clonotype", groups=ir.tl.top_n("clonotype", 10))

To discuss

• retrieve multiple columns at once / vectorization?
• what about "extra" chains?
• plotting