Request for comment: gene list functionality & data structure

Data structure

• Lists are not mutually exclusive: a gene may be present in multiple lists
• Lists are nonredundant: a gene may only be present once in a given list
• Lists are ordered: gene position in a list is meaningful
• Lists are nonexhaustive: not every gene will appear in a list
• Lists have variable length
• Lists names are strings that follow the same conventions as category names
• Feature names (i.e., genes) are present in var_names

e.g.,

{'cardiomyocyte_markers' : ['gene1', 'gene2', 'gene3', 'gene0'],
'microglia_markers': ['gene1', 'gene7, 'gene4']}

Loading gene lists

We should load gene sets from three places:
1 - adata.uns['rank_genes_groups'] These are the results of precomputed one-vs-all differential expression, and include the log fold change and adjusted p-values.

Input: structured arrays as described here
Output: gene set with title name vs. all, contents are gene histograms as rendered for differential expression (i.e., include log fold change and p-values), ordered by log fold change

2 - any .uns field specified at the command line
--gene-sets-field [Default: None] Name of field in adata.uns to look for gene set definitions; expects a dictionary of lists or a dictionary of sets.

Input: dictionary of lists, tuples, or sets. Aligns with scanpy specification of marker genes. Output: gene set with the title key, contents are gene histograms rendered in the order they were provided (if list or tuple) or alphanumerically sorted (if set)

• from a CSV that mirrors the cell annotations data format
  Example prepared here

Interacting with gene lists

Core functionality:

• Add and remove genes from lists
• Color by & select by overall expression of all genes in a list (method TBD)
• Duplicate existing lists (similar to existing cell annotation affordances)
• Move and copy individual genes between lists (e.g., drag and drop)

Saving gene lists

Still noodling on this; open questions
1 - JSON (easier merge and more ‘correct’) vs CSV (easier to edit, ship around and deal with)
Current thinking: CSV 2 - Autosave to separate file as for cell annotations? Current thinking: yes!! 3 - Mutable vs nonmutable groups as for annotations?
Current thinking: ???

Implementation plan

MVP -> full CRUD cycle

[liaprins-czi]

For reference, here is my first thought for how to indicate in the UI when listed gene list genes are not available:

[signechambers1]

Closing, tracking implementation work here