Feature: Multi-level annotations

[aichander]

Annotations following a biologically meaningful hierarchy where subtypes are nested in a parent class. This way you'll have less and more granular labels making it easier to go from annotations to analysis.

Example

• T cells
  -- CD 4 T cells
  -- CD 8 T cells
• B cells

[aichander]

def generalizer(label):
    label = re.sub(r'\(.*?\)', '', label) 
    label = label.replace('-', ' ')       
    label = label.lower()                
    label = label.replace('cell ', 'cell ')  
    return re.sub(r'\s+', ' ', label).strip()

Heres a function I use now to get to a "Level 2" state of granularity based on how the annotations are currently structured.

[parashardhapola]

We have made some progress in this direction in the latest release. The cluster clusters are now grouped by their general cell types. This is dont dynamically on each dataset.

In the UI, we show these on the sidebar:

I have updated the Colab notebook to show how to pull in this information:
https://colab.research.google.com/drive/1aRLsI3mx8JR8u5BKHs48YUbLsqRsh2N7?usp=sharing

Example dotplot with grouping information (code in the notebook above)