1

I am trying to save a file as to h5ad format and it is giving this value error; ValueError: '_index' is a reserved name for dataframe columns.

import pandas as pd
import scanpy as sc
import numpy as np
data = sc.read_h5ad('f.h5ad')
annotation = pd.read_csv('n.tsv', sep='\t')
annotation_dict = {item['barcodes']:item['celltype'] for item in annotation.to_dict('records')}
data.obs['barcodes'] = data.obs.index
data.obs['celltype'] = data.obs['barcodes'].map(annotation_dict)

sc.pp.filter_genes(data,min_cells=686)
sc.pp.filter_cells(data,min_genes=10)
sc.pp.normalize_per_cell(data,20000)
sc.pp.log1p(data)
sc.pp.highly_variable_genes(data,n_top_genes=1000)
data.X = np.exp(data.X.toarray())-1
data=data[:,data.var['highly_variable']]
sc.pp.normalize_per_cell(data,3800)

clustered = sc.read_h5ad('f.h5ad')
sc.pp.filter_cells(data,min_genes=10)
sc.pp.recipe_zheng17(clustered)

sc.tl.pca(clustered, n_comps=50)
sc.pp.neighbors(clustered, n_pcs=50)
sc.tl.louvain(clustered, resolution=0.15)
clustered.obs.groupby('louvain').count()
data.obs['louvain'] = list(clustered.obs['louvain'])
split = pd.DataFrame(data.obs['barcodes'])
test = split.sample(frac=0.2)
d_split = {item:'test' for item in test['barcodes']}
data.obs['split'] = data.obs['barcodes'].map(d_split).fillna('train')
data.write_h5ad(e.h5ad')
user3635401
  • 11
  • 4
  • Please edit your question and include the full traceback of your exception. It's difficult for us to help you if we don't know which line of your code the error is encountered on. – Luke Woodward Dec 05 '21 at 13:32

1 Answers1

1

This is probably related to a known issue with the AnnData .raw object. Two workarounds (From here):

#1
data.__dict__['_raw'].__dict__['_var'] = data.__dict__['_raw'].__dict__['_var'].rename(columns={'_index': 'features'})
 
#2, deleting the backed up raw information
del data.raw
YotamW Constantini
  • 400
  • 2
  • 11