Duplicated cell names after concatenation

In DropSeq experiments cell names are encoded by 12nt barcodes. It seems that no name check is performed when merging multiple datasets in ScanPy.

>>> from collections import Counter
>>> import scanpy.api  as sc

>>> f = sc.read("data1.txt").transpose()
>>> g = sc.read("data2.txt").transpose()
>>> c = f.concatenate(g)

>>> len(c.obs_names)
7932
>>> len(set(c.obs_names))
7890

>>> cc = Counter(c.obs_names)
>>> cc.most_common(10)
[('AAAAAAAAAAAA', 2), ('TCCTGTCTCTTA', 2), ('CGCAAGGGAAAG', 2), ('ACCCGTCTATGT', 2), ('CTCCTGTCTCTT', 2), ('TTCCTGTCTCTT', 2), ('CCCTGTCTCTTA', 2), ('CCGCTGTCTCTT', 2), ('GACAAACCTACC', 2), ('ACACTGTCTCTT', 2)]

[falexwolf]

Hi Davide,

thank you! Currently, we use the default behavior of pandas concatenate: see here. I’ll not be able to fix this during the next days. @flying-sheep, could you have a look and maybe figure out a meaningful option or meaningful default to circumvent this? It should be easy to simply pass an option to DataFrame.concat().

@dawe Thanks again for your pull request. I also put a new anndata version that incorporates it on PyPI.

Cheers, Alex

[falexwolf]

So, this is solved in anndata 0.5 and scanpy 0.4.3. See the release notes (https://scanpy.readthedocs.io) and https://github.com/theislab/anndata/commit/63500075e926f202e856bd04ec673df55bbd2460 and the example.

Hope this is a meaningful default. If you pass index_unique=None, then it keeps the previous indices including duplicates.