Fix default number_of_noise_barcodes for hashsolo

[LustigePerson]

It seems like the last PR #1483 1504a36 by @njbernstein broke the way the number_of_noise_barcodes worked.

Like in the original solo repository the number_of_**non**_noise_barcodes should be 2 as a default OR the number_of_barcodes - number_of_noise_barcodes.

However, changing the default number_of_noise_barcodes from None to 2 and changing:

number_of_non_noise_barcodes = (
        num_of_barcodes - number_of_noise_barcodes
        if number_of_noise_barcodes is not None
        else 2
    )

to

number_of_non_noise_barcodes = num_of_barcodes - number_of_noise_barcodes

led to the situation, that the used number_of_non_noise_barcodes was number_of_barcodes - 2 as a default instead of 2.

This leads to problems with demultiplexing, especially with multiple barcodes. Hence, I changed this back.

I also modified the check for a valid input to cope for the None default.

if number_of_noise_barcodes >= len(cell_hashing_columns):

changed to

if (number_of_noise_barcodes is not None) and (
        number_of_noise_barcodes >= len(cell_hashing_columns)
    ):

The rest of changes was introduced by black formatting.

[codecov]

Codecov Report

Merging #2190 (e7ae2d5) into master (4322e82) will not change coverage.
The diff coverage is 100.00%.

@@           Coverage Diff           @@
##           master    #2190   +/-   ##
=======================================
  Coverage   71.95%   71.95%           
=======================================
  Files          98       98           
  Lines       11537    11537           
=======================================
  Hits         8301     8301           
  Misses       3236     3236           

Impacted Files	Coverage Δ
scanpy/external/pp/_hashsolo.py	89.18% <100.00%> (ø)

[njbernstein]

@LustigePerson yup this was my bad. Thanks for fixing this. much appreciated

[Zethson]

Thank you!

Would it be possible to catch and validate this behavior with a test here https://github.com/theislab/scanpy/blob/master/scanpy/tests/external/test_hashsolo.py ?

[ivirshup]

@LustigePerson, would you be able to add a quick test here? Then this could get into the next release

[LustigePerson]

I just increased the noise in the testdata. With this, the assertion fails with the wrong number of noise barcodes.
The original test data was to "clean".

[ivirshup]

LGTM, thanks for the fix!