QIIME2-2017.2 installation


I am trying to install q2-2017.2 as this is the version that gneiss currently supports. However, when I try to install it using conda, it says that some of the qiime2 plugins are missing from the conda channel and upon checking, it does seem so. For example https://conda.anaconda.org/qiime2/linux-64/q2-types-2017.2.0-py35_0.tar.bz2 doesnt exist


Hi @Aditya_Bandla,

We don’t support older release while we are in alpha, so the installation instructions for 2017.2 broke when we moved the packages behind an Anaconda label.

Fortunately, gneiss is tested with the latest builds of QIIME 2, and so it should work just fine with 2017.5 (the current release). You should be able to just follow the installation instructions generally, and then install gneiss on top.

@mortonjt does that sound correct to you?

QIIME2-2017.4 installs without issues. When I try install gneiss under that using conda, I get the following errors

PackageNotFoundError: Dependencies missing in current linux-64 channels: 
  - gneiss 0.3.2* -> q2-composition ==2017.2.0
  - gneiss 0.3.2* -> q2-feature-table ==2017.2.0
  - gneiss 0.3.2* -> q2-types ==2017.2.0
  - gneiss 0.3.2* -> q2cli ==2017.2.0
  - gneiss 0.3.2* -> qiime2 ==2017.2.0

When I try to install q2-2017.5, I always end up with this error when running qiime --help at the end of the install

ImportError: C extension: No module named 'pandas._libs.tslib' not built. If you want to import pandas from the source directory, you may need to run 'python setup.py build_ext --inplace --force' to build the C extensions first.

After downgrading pandas to 0.19.1, it throws up

ImportError:lib/python3.5/site-packages/biom_format-2.1.5-py3.5-macosx-10.6-x86_64.egg/biom/_filter.cpython-35m-darwin.so: invalid ELF header

@Aditya_Bandla, how are you installing gneiss, and where from (e.g. conda, pip, source)?

That error looks like you are on a linux machine trying to use a mac binary, which won't work. There's two files on the install guide, you probably need the linux one.

Hi Evan

I used conda for the installation


Assuming you are getting it from the biocore channel, it looks like its recipe is stale.

You may be able to install from source, @mortonjt is that advisable and if so, how would one go about doing that?

Also, were you able to install QIIME 2 2017.5 successfully or are there still issues there?

Exactly, we are trying to push for the recipe to be updated with the newest qiime2 release in the next few days.

@Aditya_Bandla the conda install won’t work for gneiss with the newest version of qiime2. But you should be able to install the dev version of gneiss, which is compatible with the newest version of qiime2. Can you try

pip install git+https://github.com/biocore/gneiss.git

Gneiss is released from the biocore channel now and can be installed alongside QIIME 2 2017.6!

From within your QIIME 2 environment run:

conda install -c biocore gneiss

Hi Evan

qiime2 (since release 5) installation seems to be failing with the following error. version 4 however seems to work

ImportError: /home/users/nus/a0113466/scratch/miniconda3/envs/qiime2-2017.6/lib/python3.5/site-packages/biom_format-2.1.5-py3.5-macosx-10.6-x86_64.egg/biom/_filter.cpython-35m-darwin.so: invalid ELF header

Hi @Aditya_Bandla, it looks like you might accidentally be using the Mac OS X installation commands, instead of the linux installation commands (as @ebolyen mentioned above). If you are trying to install QIIME 2 on a linux machine, you will need to use the linux installation instructions. So ultimately this would look like the following, to get the latest versions of QIIME 2 and gneiss:

# These first two might be optional, depending on how far you got earlier
$ source deactivate qiime2-2017.6
$ conda env remove -n qiime2-2017.6
# The following steps are not optional
$ conda create -n qiime2-2017.6 --file https://data.qiime2.org/distro/core/qiime2-2017.6-conda-linux-64.txt
$ source activate qiime2-2017.6
$ conda install -c biocore gneiss

Hi Ryan

I double checked to make sure that I was using the linux files, when installing via conda. Also, qiime2-2017.7 doesnt seem to currently have a conda recipe?

Hi @Aditya_Bandla, I wasn't able to determine from your message if you were able to successfully follow the instructions I provided above --- the errors you provided us indicate that you are using the Mac OS installation files instead of the linux installation files. I did just realize that you might need to clean up any of your corrupted mac os-based environments first, I have updated the code block above to reflect this.

We haven't release QIIME 2 2017.7 yet (it isn't slated to be released until July 2017 --- that is what the version number means: YYYY.MM), but stay tuned! QIIME 2 2017.6 is the current version of QIIME 2.

Hi Ryan

I have been working with the qiime2 alpha releases for quite a while now, using the conda linux recipes. I am working on a linux cluster, so I dont have any mac based conda environments

Hence my comment that I have been having this error since version 5. I do realize that somehow the mac version of the biom format package is getting downloaded. But this seems to happen when I use the linux recipe

Anyways I found a way resolve this. Thanks for your support


Hi @Aditya_Bandla!

We are in agreement!

I don't think this is correct --- you appear to have accidentally installed the mac version of qiime2 on your linux cluster.

Again, this appears to be related to a copy-paste error: the reason the wrong biom package was downloaded was because the wrong version of qiime2 was installed.

I suggest you double-check which recipe you are using, the errors you provided indicate that you used the mac os recipe (probably on accident --- the mac os instructions are first on the installation page, it is easy to copy the wrong command). The example I provided above should be a quick minimal set of commands for getting rid of the mac os environment that was accidentally installed and installing the correct linux version. Good luck!


1 Like

This topic was automatically closed 31 days after the last reply. New replies are no longer allowed.