Installation problem - packages conflicts

[P D]

Hi there,

I am struggling with installing iphop.

I've tried to install the environment from yml file but also from bioconda, on WSL. I also tried to install it on HPC and that didn't work either.

In each case incompatible packages were found.

The log is attached.

[Simon Roux]

Hi,

iPHoP has not been tested on WSL, so I can’t guarantee that it should work (unfortunately WSL is not, in my experience, a fully functional equivalent to a linux system). I am more surprised by the fact that you could not install iPHoP via conda on an HCP system. Do you know what your OS is there ? As far as I know, installing it with mamba from a “clean” environment has worked for most people so far (“mamba create -c bioconda -c conda-forge -n iphopenv iphop”).

Best,

Simon

[P D]

‌

Hello Simon,

I would like to apologize for late response, but I was on my leave.

I am surprised, because I have not faced any problems with WSL so far. As you suggested, I tried to install iphop from a clean environment on the HPC cluster and it did not work as well. Please find the log attached. Furthermore, I tried to install iphop on freshly installed Ubuntu 22.04 on my laptop and I keep getting exactly the same error like in the attached log. It seems that biopython is causing those issues.

Moreover, the docker container converted to a singularity image is not working on the HPC system, but it does work as a normal docker container on Windows 11. Unfortunately, I do not have any PC with sufficient RAM memory to run the software with docker.

The log from the singularity is as follows: FATAL: "iphop": executable file not found in $PATH

The Linux distribution installed on the cluster: Rocky Linux release 8.7 (Green Obsidian).

In conclusion, unfortunately I do not see any other way to run the software.

All the best!

[Simon Roux]

I am genuinely confused, I was just able to install iPHoP on a linux cluster working from Rocky 8.8 (Green Obsidian) as well as on a Ubuntu 22.04, using essentially the same command line. The only difference was my version of mamba was 1.1 vs 1.4 for you, but I’m not sure that it should cause that many issues. I am wondering now if this is a question of python version..

Maybe one thing to try would be the following:

$ conda create -n iphop_env -c conda-forge mamba python=3.8
$ conda activate iphop_env
$ mamba install -c bioconda -c conda-forge iphop

This is not ideal because it sort of creates a conda/mamba within another conda/mamba, but may be worth trying ?

‌

[P D]

I am afraid but this solution is not working.

However, you may be right that this is the case with the python version.

Literally, I had the same problem with python, perl, biopython and their dependencies several months ago, and it was caused by the recent python update 3.9 → 3.10 (some packages consider python 3.10 as 3.1). However, even installing python 3.8 did not resolve the problem here.

I tried to track down the malfunctioning packages and manually install them. In the middle of the list, some packages were to be updated and some to be downgraded, so it was a meaningless loop. For example, biopython=1.79, which caused incompatibilities before, can be installed manually, but later on its dependencies have to be updated/downgraded because of perl.

‌

[Simon Roux]

Right, you may be right that the issues you see may be related to the version of python/perl installed system-wide, which then conditions everything downstream (conda version, etc). It is still a bit strange to me that you see the same issues in every system, including e.g. Ubuntu 22.04, while it seems to work for most users (including the automated build via bioconda). If there is any possibility, it seems like maybe a cloud computing instance could be a solution for you ? (that’s the way bioconda test their package, and it seems to work with default instances).