Local pipeline install fails during convert_matrix_from_mtx_to_loom.py

[bimsbbioinfo]

Local pipeline install fails during convert_matrix_from_mtx_to_loom.py

[t-carroll]

Was able to troubleshoot this further. Looks like in convert_matrix_from_mtx_to_loom.py, there is a block of code:

    path_genes    = os.path.join(path_input,'features.tsv')
    genes         = pd.read_csv(path_genes, sep='\t', header=None)
    genes.columns = ['gene_id','gene_id2']
    genes         = genes.drop(columns = ['gene_id2'])

Where it is expecting 2 columns in the "features.tsv" file. However, all of the relevant features.tsv files generated by the test pigx_scrnaseq command in my environment generate 3 column outputs. Here's an example, but it's the same for all the various features.tsv generated for each sample:
head -3 tests/out/Mapped/WT_HEK_4h_br1/hg19/WT_HEK_4h_br1_Solo.out/Gene/filtered/features.tsv

ENSG00000080573 ENSG00000080573 Gene Expression
ENSG00000267650 ENSG00000267650 Gene Expression
ENSG00000080511 ENSG00000080511 Gene Expression

I therefore modified the relevant code on my fork's devel branch to the following:

    path_genes    = os.path.join(path_input,'features.tsv')
    genes         = pd.read_csv(path_genes, sep='\t', header=None)
    genes.columns = ['gene_id','gene_id2','gene_expression']
    genes         = genes.drop(columns = ['gene_id2','gene_expression'])

And the test command now completes without error. Do the features.tsv files generated by your guix install on the test command only have two files? If so I imagine I have the wrong version of some dependency (maybe STAR? I have 2.7.5a).

In any case, the HTML report generated by my test command (attached) seemingly indicates I am having further issues with the pipeline, and may be generating incorrect results. For instance, the number of uniquely mapped UMIs is a tiny fraction of the total UMIs, and plots 3.0.2 and 3.0.3 are completely blank. Would it be possible to generate an example report for your test pipeline command (e.g. hg19.scRNA-Seq.report.html generated by your own installation) for users to compare against to insure proper installation?

Thanks for the help!
hg19.scRNA-Seq.report_files.zip

[frenkiboy]

Dear Tom,

Thank you so much for reporting the message. I'll look into the error and sort it out.

Could you please just tell me which pandas version you are using?

Best,

Vedran

[t-carroll]

Hi Vedran,

Thanks for looking into it! It looks like I'm using pandas version 1.0.3:

pip freeze | grep "pandas"

pandas==1.0.3