What is it about?
Estimating a phylogenetic tree from sequence data is a five-stage pipeline: fetch sequences, find homologs, align, trim, then infer the tree. On a SLURM or PBS cluster the same pipeline also means no root installs, a queue, and a shared filesystem that punishes stray I/O. Web servers hide the parameters. Ad hoc scripts run on one login node and nowhere else. This paper is a literate workflow for that pipeline, demonstrated on a salmonid protein set. Alignment uses MAFFT or MUSCLE-5, trimming uses Gblocks, and the tree is RAxML-NG. The environment is a micromamba stack from BioConda, recorded in environment.yml, not Nix, because the R analysis needs those packages. Quarto holds the executable document. radian and httpgd give an interactive R session on the login or compute node. ggtree draws the production figures from the same document.
Featured Image
Photo by Google DeepMind on Unsplash
Why is it important?
A point-and-click phylogeny server will not scale, and it will not record the MAFFT versus MUSCLE-5 choice that later changes the tree. A cluster without a recorded environment will not rebuild next year. The claim is the combination: a BioConda lock for the tools, a Quarto document for the parameters and the figures, and an interactive path that does not require the user to become a queue administrator. The biology is the example, not the result. The same pattern (fetch, align, infer, plot, all in one literate file on a queue) applies to any multi-stage life-science workflow that already has BioConda packages.
Perspectives
The biologists I was working with needed trees on the national cluster and had no interest in writing module files. I packaged the tools with micromamba, not Nix, because BioConda already had MAFFT, MUSCLE-5, RAxML-NG, and the R stack. Quarto plus radian was the piece that kept every parameter next to the figure it produced. The salmonid subset is there so the document is a real analysis, not a toy. I left the evolutionary-model debate out of the paper on purpose. The point was the HPC practice: an environment.yml and a Quarto book a biologist can re-run.
Rohit Goswami
University of Iceland
Read the Original
This page is a summary of: High Throughput Reproducible Literate Phylogenetic Analysis, November 2022, Institute of Electrical & Electronics Engineers (IEEE),
DOI: 10.1109/pdgc56933.2022.10053210.
You can read the full text:
Contributors
The following have contributed to this page
