Add anndata factory #255
Add anndata factory #255
Conversation
There was a problem hiding this comment.
Yeah 🥳 Fantastic!
- I think how to handle duplicated protein groups is a good question - is this expected to happen? Otherwise I would raise a warning/error, drop them, and use the strategy
firstwhile pivoting - For some downstream analyses it might be good to consider additional information from the psm files (e.g. gene names). Would it be possible to add additional metadata to the metadata attributes? (e.g. list of columns the
.obsand.obsattributes?)
alphabase/anndata/anndata_factory.py
Outdated
| index=PsmDfCols.RAW_NAME, | ||
| columns=PsmDfCols.PROTEINS, | ||
| values=PsmDfCols.INTENSITY, | ||
| aggfunc=np.nanmean, # how to aggregate intensities for same protein in same raw file TODO first? |
There was a problem hiding this comment.
Are there scenarios in which the same protein occurs multiple times in a file? I tested the diann_test_input_mDIA.tsv with the DiannReader class and did not find any.
I think aggregating by the mean might be dangerous. One could add a test on whether there are duplicates, and at least raise a warning.
duplicated_proteins = self._psm_df[PsmDfCols.PROTEINS].duplicated()
if duplicated_proteins.sum() > 0:
warning.warn(f"{duplicated_proteins.sum()} duplicated protein groups") Alternatively, this could be an optional argument agg_duplicates: Literal["mean", "drop", "raise"] with "raise" raising a ValueError, "drop" dropping the duplicated entries, and "mean" aggregating
There was a problem hiding this comment.
user-define agg_duplicates => https://github.com/orgs/MannLabs/projects/20/views/1?pane=issue&itemId=88563720
| if missing_cols: | ||
| raise ValueError(f"Missing required columns: {missing_cols}") | ||
|
|
||
| self._psm_df = psm_df |
There was a problem hiding this comment.
Would it be possible to add optional metadata columns to the .obs and .var attributes by passing obs_columns: Optional[str, List[str]] and var_columns: Optional[str, List[str]] to the factory class?
This would add to the complexity as one had to validate that the columns are in the data frame, but other than that one could just use .pivot_table while passing the list of columns
There was a problem hiding this comment.
mschwoer
force-pushed
the
add_anndata_factory
branch
from
36f0ce4 to
c19e9ce
Compare
| @@ -1,3 +1,4 @@ | |||
| anndata==0.11.1 | |||
There was a problem hiding this comment.
the problem is that in order to use 0.11.1 we would need to drop support for python 3.8 (latest supported versions are 0.10.9, 0.11.0rc1, 0.11.0rc2) .. would be an argument for moving this module out of alphabase
mschwoer
force-pushed
the
add_anndata_factory
branch
from
f6bb454 to
427e64a
Compare
mschwoer
changed the base branch from
refactor_readers_XI
to
add_alphadia_reader
mschwoer
force-pushed
the
add_alphadia_reader
branch
from
3e745b3 to
239e2ef
Compare
mschwoer
force-pushed
the
add_anndata_factory
branch
from
427e64a to
8ce3ce5
Compare
|
Check out this pull request on See visual diffs & provide feedback on Jupyter Notebooks. Powered by ReviewNB |
Add first version of anndata conversion.