feat(downsample):AnnData input and output for downsample_cells

[Chloe-Thangavelu]

Summary:
Modified downsample_cells function to accept anndata.AnnData objects as input. When an AnnData object is provided, .X and .obs data are combined into a pandas DataFrame, before applying the rest of the downsampling function.

Changes:

• Added code to convert anndata objects into pandas dataframes
• Returns error message if input is neither an anndata object or pandas dataframe
• Created unit test to check anndata object is correctly processes

[Chloe-Thangavelu]

Summary:
Modified downsample_cells function to accept and return anndata.AnnData objects. When an AnnData object or pandas dataframe is provided, annotation information is extracted and used in a helper function to determine which cell indexes to keep.

Changes:

• Added code to handle two different input data types: annData object or pandas dataframe
• Placed downsampling logic into a helper function that returns which cell ID indexes to keep.
• The input data are subset according to these returned indexes in the main function.
• Created unit test to check anndata objects are correctly processed and returned

[Copilot]

Pull Request Overview

This PR updates the downsample_cells function to support input as an AnnData object, converting its .X and .obs into a DataFrame for downsampling, while maintaining compatibility with pandas DataFrames.

• Added conversion logic for AnnData input
• Updated error handling and documentation for input types
• Introduced a new unit test to validate AnnData processing

Reviewed Changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
tests/test_data_utils/test_downsample_cells.py	Added a new unit test to ensure downsample_cells correctly processes AnnData objects
src/spac/data_utils.py	Refactored downsample_cells logic to handle both pandas DataFrame and AnnData inputs, updating docstrings and internal variable usage

[Copilot]

The repeated call to logging.basicConfig in both downsample_cells and _get_downsampled_indexes may cause configuration conflicts; consider configuring logging once at application startup.

Suggested change

	logging.basicConfig(level=logging.WARNING)
	# Convert annotations to list if it's a string
	if isinstance(annotations, str):
	annotations = [annotations]
	# Check if the columns to downsample on exist
	missing_columns = [
	col for col in annotations if col not in input_data.columns
	]
	if missing_columns:
	raise ValueError(
	f"Columns {missing_columns} do not exist in the dataframe"
	)
	# If n_samples is None, return the input data without processing
	if n_samples is None:
	return input_data.copy()
	# Combine annotations into a single column if multiple annotations
	if len(annotations) > 1:
	# Combine annotations into a single column if multiple annotations
	if len(annotations) > 1:

[Copilot]

For consistency and clarity, update the error message to refer to 'AnnData' (with proper casing) instead of 'Anndata'.

Suggested change

	else:
	raise TypeError("Input data must be a Pandas DataFrame or Anndata Object.")
	else:
	raise TypeError("Input data must be a Pandas DataFrame or AnnData Object.")

[fangliu117]

The "combined_col_name" parameter is documented but never used in the code.
Since grouping_col is a pd.Series and not a new column in the cell_data DataFrame, the combined_col_name parameter isn't strictly necessary. May either remove or assign the name to the Series.

[fangliu117]

The apply method for combining annotations is readable but can be slow on very large datasets. A more performant, vectorized approach is to use str.cat or agg.

grouping_col = cell_data[annotations].astype(str).agg('_'.join, axis=1)
(Ensure all columns are string type first)

[fangliu117]

The variable grouping_col is used inconsistently.
Suggested:
if len(annotations) > 1:
cell_data[combined_col_name] = cell_data[annotations].apply(
lambda row: '_'.join(row.values.astype(str)), axis=1)
grouping_col = combined_col_name
else:
grouping_col = annotations[0] # This is a string, not a Series

[Chloe-Thangavelu]

Thank you, I will go through & add these suggestions today

[fangliu117]

The test calls the function with stratify=False and min_threshold=5. In the downsample_cells function, the min_threshold parameter is only used when stratify=True.

[Chloe-Thangavelu]

The changes should be complete now. I have tested the code and it looks like its working. Let me know if any other fixes need to be made.