Add rammap as a supported read mapping program#297
Conversation
rammap (https://github.com/jwanglab/rammap) is a pure-Rust, minimap2-compatible read mapper available on Bioconda. Wire it in as a new `--mapper rammap` choice alongside the existing mappers. - Add MappingProgram::RAMMAP and handle it in every mapping-program match (command construction, index maintenance, parameter selection). - Map short reads with `rammap -x sr -a`, matching CoverM's default short-read use case, and expose `--rammap-params` for extra options. - Add check_for_rammap dependency/version check. - Register rammap in the CLI mapper list, help table, README dependency list, and pixi test environment; add a `make` smoke test. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01D4S2TfmfT9BsYgJF5VDz81
rammap, like bwa-mem2 and strobealign, is only installed on linux-64 (it is not available for osx via bioconda). The macOS CI jobs already skip the bwa_mem2 and strobealign tests for this reason; add rammap to that skip list so test_make_rammap does not fail for a missing executable. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01D4S2TfmfT9BsYgJF5VDz81
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b6170576b1
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| bwa-mem2 = "*" | ||
| dashing = ">=0.4.0,<1.0" # dashing 1.0 build gives illegal instruction errors | ||
| strobealign = ">=0.14.0" # tests generate the strobealign index at runtime | ||
| rammap = ">=1.1.1" |
There was a problem hiding this comment.
Line 31 adds the new dependency, but pixi.lock still has no rammap entry (rg rammap pixi.lock). The locked Linux workflow runs pixi run --frozen cargo test in .github/workflows/test-coverm.yml line 28, and Pixi documents --frozen as installing from the lockfile without updating it, so the new test_make_rammap runs in an environment without the binary and check_for_rammap fails. Please regenerate and commit pixi.lock with rammap included.
Useful? React with 👍 / 👎.
| } | ||
|
|
||
| #[test] | ||
| fn test_make_rammap() { |
There was a problem hiding this comment.
This test is unconditional, but the macOS workflows still run cargo test -- --skip bwa_mem2 --skip strobealign in .github/workflows/test-coverm.yml lines 87 and 120, while rammap was added only under [target.linux-64.dependencies]. In those jobs, selecting --mapper rammap reaches parse_mapping_program and check_for_rammap, so the test fails before mapping because the binary is absent. Please either add a macOS rammap dependency or skip/gate this test on platforms without rammap.
Useful? React with 👍 / 👎.
| MappingProgram::STROBEALIGN => "strobealign".to_string(), | ||
| // rammap is a minimap2-compatible aligner; map short reads with | ||
| // the 'sr' preset and emit SAM with -a. | ||
| MappingProgram::RAMMAP => "rammap -x sr -a".to_string(), |
There was a problem hiding this comment.
When CoverM users choose --single with --mapper rammap, this command still forces -x sr; in the upstream rammap CLI source, the sr preset sets FRAG_MODE, and the CLI derives pe_mode = two_file_pe || frag_mode, then reads the second mate from the same iterator when only one query file is supplied. That makes a single-end FASTQ get consumed as interleaved pairs, with an odd final read dropped, instead of being mapped as single-end reads. Please add --frag no/--pairing no for ReadFormat::Single or otherwise vary the rammap options by read format.
Useful? React with 👍 / 👎.
rammap's `sr` preset enables fragment mode, which pulls read pairs from a single input file (this is how rammap maps interleaved reads). For genuinely single-end input (`--single`) that would incorrectly consume the file as interleaved pairs and drop an odd trailing read. Pass `--frag no` for ReadFormat::Single so single-end reads are mapped individually, while coupled and interleaved input keep the default fragment-mode behaviour. Add a single-end rammap smoke test covering this path. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01D4S2TfmfT9BsYgJF5VDz81
Commit the rammap linux-64 conda package (1.1.1, build hfa8f182_0) into the frozen lockfile so `pixi run --frozen` installs it in the locked CI job, matching how strobealign and bwa-mem2 are pinned. Package hash and metadata taken from the bioconda repodata and verified against the downloaded artifact (sha256 4e03e5..., 885279 bytes). Its only non-system dependency, libgcc >=14, is already satisfied by libgcc 15.1.0 in the lock. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01D4S2TfmfT9BsYgJF5VDz81
2 participants
Summary
This PR adds support for rammap, a minimap2-compatible read aligner, as a mapping option in CoverM alongside existing mappers like bwa-mem, minimap2, and strobealign.
Key Changes
-x srpresetrammap -x sr -a(short-read preset with SAM output)--rammap-paramsCLI option to allow users to specify additional parameters, with the same security warnings as other mapping toolsMappingProgram::RAMMAP) and updated all relevant match statements across the codebasecheck_for_rammap()functiontest_make_rammap) to verify rammap mapping functionality works end-to-endImplementation Details
https://claude.ai/code/session_01D4S2TfmfT9BsYgJF5VDz81