Dunn Lab Practices

This chapter is the most opinionated in the manual, and deliberately so. Everything before it is guidance we would stand behind for anyone; this is how we have settled the questions that have more than one defensible answer.

If you are outside the lab, treat it as a worked example rather than a recommendation. The value is less in our specific choices than in the fact that they are written down and encoded somewhere a tool can apply them — the alternative is a convention that exists only in the head of whoever set it up.

Conventions

These conventions are encoded as skills in the plugin, so Claude applies them as you work — that chapter maps what is in there. What follows is the reasoning, which the skills themselves do not carry.

A few are more consequential than a style preference, and they are the ones most likely to surprise someone joining:

Python by default, R when a library requires it. We prefer industry-standard tools over domain-specific ones, as Getting Started argues. R remains the right answer when the analysis needs a package that only exists there, when it is what you know and it works, or when a collaboration has already chosen it.

Raw data is immutable, enforced structurally. data/raw/ is never written to. Every transformation produces a new file under data/processed/ from a script that can be re-run. This is a general principle — Using AI in Research makes the case — but the bioinformatics skill turns it into specific checks.

Cross-species gene IDs are namespaced as Genus_species@gene_id. The @ separator is chosen because it does not appear in standard gene IDs and is not a shell metacharacter. Every renaming keeps a mapping file, so the transformation is always reversible. Merging datasets without this is a class of silent error that surfaces months later in a tree.

Never abbreviate an author list, and never guess a bibliographic field. A missing DOI gets a % TODO comment, not a plausible-looking value. This sounds pedantic until an AI assistant fills one in for you.

CLAUDE.md stays under 100 lines. The official guidance targets 200; we hold to half that, because everything in a CLAUDE.md is paid for in every session. When project guidance outgrows it, the answer is a path-scoped rule in .claude/rules/, not a longer CLAUDE.md — see Managing Context.

Checkpointing is by output existence, not sentinel files. A pipeline stage is skipped if its output already exists. To re-run a stage you delete its output. There is no hidden state tracking what has completed.

Data management

Consult our separate data management plan for requirements and conventions on where to store data, how to share files, archival, and retention.