"protein(s) without contig" leading to "ERROR:vcontact2: Error in contig clustering"

[Former user]

Hi,

I'm trying to run vcontact2 on two phage contigs, but I keep getting "protein(s) without contig" warnings in "Building the cluster and profiles" followed by a possibly related fatal error in "Contig Clustering & Affiliation" ERROR:vcontact2: Error in contig clustering ERROR:vcontact2: 'DataFrame' object has no attribute 'ix'

A log of the run and my gene-to-genome mapping file are attached. I'm running vcontact2-0.9.13 installed via conda on an ubuntu 18.04 system.

Any help would be appreciated

[Thomas Hackl]

OK, red herring = the "protein(s) without contig" errors occur ed because I forgot the header line in the “gene-to-genome”.csv…

However, even with the proper mapping file I get:

------------------------Contig Clustering & Affiliation-------------------------
Loaded graph with 2243 nodes and 79278 edges
[====================] 100% Growing clusters from seeds...
[====================] 100% Finding highly overlapping clusters...
[====================] 100% Merging highly overlapping clusters...
Detected 319 complexes
ERROR:vcontact2: Error in contig clustering
ERROR:vcontact2: 'DataFrame' object has no attribute 'ix'

‌

[Thomas Hackl]

And apparently that can be solved by downgrading pandas (https://bitbucket.org/MAVERICLab/vcontact2/issues/17/error-vcontact2-error-in-contig-clustering).

Sometime it just helps talking out loud . Sorry for the fuss.

[Thomas Hackl]

Ok, also had to downgrade to python=3.7 for panda=0.25.3 to work. But now I get another error…

------------------------Contig Clustering & Affiliation-------------------------
Loaded graph with 2243 nodes and 79278 edges
[====================] 100% Growing clusters from seeds...
[====================] 100% Finding highly overlapping clusters...
[====================] 100% Merging highly overlapping clusters...
Detected 319 complexes
ERROR:vcontact2: Error in viral clusters
ERROR:vcontact2: type object 'object' has no attribute 'dtype'

‌

[Eugen Pfeifer]

Have the same issue by running the test data set.

ERROR:vcontact2: Error in viral clusters
ERROR:vcontact2: type object 'object' has no attribute 'dtype'

python=3.8.6 and pandas=0.25.3

Looking forward for help!

[Thomas Hackl]

What worked for me, in the end, was installing the right pandas and python first:

# python needs to be 3.7 for pandas < 1.0
conda create --name vContact2 python=3.7 # NOTE the explicit 3.7 python here
conda activate vContact2
pandas need to be 0.25.3 - https://bitbucket.org/MAVERICLab/vcontact2/issues/17/error-vcontact2-error-in-contig-clustering
conda install pandas=0.25.3
conda install vcontact2
conda install -y mcl blast diamond

‌

[Eugen Pfeifer]

Thanks a lot @Thomas Hackl!!

I reinstalled using your order and package version. Now it is working!

[Bridget Hegarty]

I’m having the same error with the example data. I tried specifying the versions of pandas (0.25.3) and python (3.7.9) like you suggested @Thomas Hackl , but I’m still getting the same error message. If anyone has any other suggestions, I’d greatly appreciate them. Thanks!

[Samuel Barnett]

In case someone is still having the “ERROR:vcontact2: Error in viral clusters, ERROR:vcontact2: type object 'object' has no attribute 'dtype'“ error, downgrading numpy to numpy=1.19.5 seemed to work for me. I’m not sure if also specifying python=3.7 and pandas=0.25.3 is also needed with this fix though because I did do that as well and continued to have the same error before downgrading numpy.

[Ben Bolduc]

I updated both the Singularity definition file and setup.py requirements to enforce numpy and pandas versions, but I haven’t yet gotten to Anaconda, so if you installed vContact2 through Anaconda, a quick fix was posted by Thomas. Indeed, using pandas <=0.25.3 and numpy <1.19.5 should resolve this issue.

I will also add a check to ensure headers are enforced in the gene-to-genome file…

[Bridget Hegarty]

thanks for the additional fixes! I appreciate the help greatly. Unfortunately, I have a new error that is identical to that in issue 36, so commented there with the new problem.

[Ben Bolduc]

Closing this as it's due to absent headers or use of .ix. A check for headers will be included in 0.9.22. Pandas versioning will be used to avoid .ix depreciation.