Missing genomes in output

[Stephen Nayfach]

I've run the latest version of the tool with the test data and it completed without errors, but some genomes are missing from the output.

‌

There are 575 user-supplied genomes in the test file:

\$ cut -f2 MAVERICLab-vcontact2-a3541dd53c3e/test_data/proteins.csv -d ',' | sed 1d | sort -u | wc -l

575

‌

But only 246 are found in the viral_cluster_overview file:

\$ grep -o 'VIR' viral_cluster_overview.csv | wc -l
246

‌

And the genome_by_genome file is not present:

\$ ls

c1.clusters modules_mcl_5.0_modules.pandas
c1.ntw modules_mcl_5.0_pcs.pandas
merged.dmnd sig1.0_mcl2.0_clusters.csv
merged.faa sig1.0_mcl2.0_contigs.csv
merged.self-diamond.tab sig1.0_mcl2.0_modsig1.0_modmcl5.0_minshared3_link_mod_cluster.csv
merged.self-diamond.tab.abc sig1.0_mcl5.0_minshared3_modules.csv
merged.self-diamond.tab.mci vConTACT_contigs.csv
merged.self-diamond.tab_mcl20.clusters vConTACT_pcs.csv
merged.self-diamond.tab_mcxload.tab vConTACT_profiles.csv
merged_df.csv vConTACT_proteins.csv
modules.ntwk viral_cluster_overview.csv
modules_mcl_5.0.clusters

‌

Any help would be appreciated

[Stephen Nayfach]

Simon Roux helped me out and found the bug causing this issue. It turns out that vcontact 2 has a problem handling distinct contig identifiers that overlap in their characters. See below for an illustrative example:

‌

My original input file looked like something this:

protein_id contig_id keywords
OTU-48322_1 OTU-48322 dummy
OTU-48322_2 OTU-48322 dummy
OTU-4832_1 OTU-4832 dummy
OTU-4832_2 OTU-4832 dummy

‌

Changing to this solved the issue:

protein_id contig_id keywords
OTU-48322_1 OTU-48322 dummy
OTU-48322_2 OTU-48322 dummy
OTU-04832_1 OTU-04832 dummy
OTU-04832_2 OTU-04832 dummy

So in the original input file having both OTU-4832 and OTU-48322 was problematic for some reason.

‌

‌

[Ben Bolduc]

Thank you for the feedback, and thanks a bunch for figuring out what caused it. I thought an earlier update resolved all the “leaky contigs” from the pipeline, but it appears I was incorrect. I have identified the offending line but need to work out the patch details.

Thanks again!

[Ben Bolduc]

On hold until fix released. In the meantime, a workaround is to ensure contig names are not subsets of each other.

[Ben Bolduc]

Fix implemented in 0.9.9. Please update to this version (or later).

Please re-open if the problem persists in a recent version! Thanks!