Use container solution for covviz installation instead of conda

[ManavalanG]

covviz is currently not available via bioconda and so we have been using pip to install it within conda environment. Current config:

name: covviz

channels:
  - conda-forge

dependencies:
  - python==3.8.5
  - pip
  - pip:
      - covviz==1.2.2

Using pip within conda env is not recommended at all though. When dealing with #49, it became clear that we needed to move away from this solution.

Since we just need covviz and no other tools need to be installed, we can sidestep conda env and instead we can directly use covviz docker container provided by the tool authors. Note that, we are already using similar setup to install and use somalier.

[ManavalanG]

So this raises the question of why use conda to install tools at all and why not instead use singularity directly. My current thoughts:

• When all the tools are available via conda package manager
  • When there are >1 tool, use conda env. Singularity would not be a readymade solution here.
  • When only one tool needs to be installed, it may be a good idea to just use the singularity solution instead of conda env.
    • If tool is in bioconda, then this tool should also be readily available via Biocontainers.
    • If not, check if tool authors made it available via docker/singularity registry.
    • If it is a python tool, installation via conda+pip should be the final resort.
• If the tool is not available via conda package manager
  • See if they are available via docker/singularity registry
  • See if you can build a singularity container, host it somewhere central to lab and then use it in your project
  • Final resort: Install the tool in local machine and use it. Try to avoid this as much as possible.

[ManavalanG]

This solution could be applied to rules using samtools, qualimap, etc. Need to discuss and double check on the pros/cons. This is not high priority though.

[ManavalanG]

mentioned in commit 92db1d1