add custom node to do pseudobulk

[jessegmeyerlab]

The following works in my python environment, return of pseudobulk data not required but we need some mechanism to output tables that are created through such code

`
import scanpy as sc
import pandas as pd
from scipy.stats import ttest_ind
from statsmodels.stats.multitest import multipletests

def pseudobulk_differential_expression(adata, leiden_group, group1, group2):
# Subset the data to include only the cells from the specified Leiden cluster
adata_subset = adata[adata.obs['leiden'].isin(leiden_group)].copy()

# Create a new column combining 'group' and 'rep' for aggregation  
adata_subset.obs['group_rep'] = adata_subset.obs['group'].astype(str) + '_' + adata_subset.obs['rep'].astype(str)  
  
# Aggregate to create pseudobulk profiles  
pseudobulk = adata_subset.to_df().groupby(adata_subset.obs['group_rep']).mean()  
  
# Get unique combinations of group and rep  
group_reps = adata_subset.obs['group_rep'].unique()  
  
# Split the pseudobulk profiles by the biological groups  
pseudobulk_group1 = pseudobulk.loc[[gr for gr in group_reps if group1 in gr]]  
pseudobulk_group2 = pseudobulk.loc[[gr for gr in group_reps if group2 in gr]]  

# Perform t-tests for each gene  
p_values = []  
fold_changes = []  
for gene in adata.var_names:  
    # Calculate fold change  
    fold_change = pseudobulk_group1[gene].mean() - pseudobulk_group2[gene].mean()  
    fold_changes.append(fold_change)  
      
    # Perform t-test  
    t_stat, p_val = ttest_ind(pseudobulk_group1[gene], pseudobulk_group2[gene], equal_var=False)  
    p_values.append(p_val)  
  
# Adjust p-values for multiple testing using BH correction  
bh_corrected_pvals = multipletests(p_values, method='fdr_bh')[1]  # what is a less conservative correction?

# Create a DataFrame with the results  
results_df = pd.DataFrame({  
    'gene': adata.var_names,  
    'fold_change': fold_changes,  
    'p_value': p_values,  
    'bh_corrected_p_value': bh_corrected_pvals  
})  
  
# Return the results DataFrame  
return results_df.sort_values('bh_corrected_p_value'), pd.concat([pseudobulk_group1, pseudobulk_group2] )

`