feat: Rp-Bp opt-in ORF caller

CHANGELOG.md

@@ -25,6 +25,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - [#165](https://github.com/nf-core/riboseq/issues/165) - Add `--extended_orf_analysis` (default `false`) which routes the hybrid GTF into the genome-BAM ORF callers (Ribo-TISH `predict`, Ribotricer `prepare-orfs`) so novel intergenic ORFs are within scope. RiboCode and transcriptome-BAM consumers stay on the canonical backbone. When the flag is set without a novel-transcript source, the pipeline warns and falls back to canonical ([@pinin4fjords](https://github.com/pinin4fjords))
 - [#171](https://github.com/nf-core/riboseq/issues/171) - Under `--extended_orf_analysis true`, run a second STAR pass for Ribo-seq samples against the hybrid transcriptome so RiboCode can call ORFs on novel transcripts; RiboCode then consumes the hybrid transcriptome BAM and hybrid GTF. Outputs land under `<outdir>/hybrid_star/`; the default-off path is unchanged ([@pinin4fjords](https://github.com/pinin4fjords))
 - [#170](https://github.com/nf-core/riboseq/issues/170) - Add PRICE (Erhard et al. 2018) as an opt-in ORF caller via `--run_price` (default `false`). Invoked one-shot across the riboseq cohort (`gedi -e IndexGenome` + `gedi -e Price`); calls flow into the cross-caller ORF catalogue. Container via `bioconda::gedi=1.0.6a` ([@pinin4fjords](https://github.com/pinin4fjords))
+- [#169](https://github.com/nf-core/riboseq/issues/169) - Add Rp-Bp (Malone et al. 2017) as an opt-in ORF caller via `--run_rpbp`. Implemented via the upstream `fasta_gtf_bam_rpbp` subworkflow (per-tool processes for extract-metagene-profiles, estimate-metagene-profile-bayes-factors, select-periodic-offsets, extract-orf-profiles, estimate-orf-bayes-factors, select-final-prediction-set, plus a shared prepare-genome). Honours `--extended_orf_analysis` by feeding the hybrid GTF when active. Per-sample predicted-ORF BED + DNA + protein FASTA under `<outdir>/orf_predictions/rpbp/`. Expect ~20-24h per replicate at genome-wide scale; pipeline emits a runtime warning when enabled ([@pinin4fjords](https://github.com/pinin4fjords))
 
 ### `Fixed`
 
@@ -82,6 +83,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 |                     | `--run_price`                            |
 |                     | `--extra_price_indexgenome_args`         |
 |                     | `--extra_price_price_args`               |
+|                     | `--run_rpbp`                             |
+|                     | `--extra_rpbp_preparegenome_args`        |
+|                     | `--extra_rpbp_predictorfs_args`          |
 
 ### `Dependencies`
 

conf/modules.config

@@ -1266,4 +1266,20 @@ process {
         ]
     }
 
+    withName: 'RPBP_PREPAREGENOME' {
+        ext.args   = { params.extra_rpbp_preparegenome_args ?: '' }
+    }
+    withName: 'RPBP_ESTIMATEMETAGENEBAYESFACTORS|RPBP_ESTIMATEORFBAYESFACTORS' {
+        // Stan MCMC, 20-24h per replicate at genome scale; first-attempt headroom.
+        time       = { 30.h * task.attempt }
+    }
+    withName: '.*:FASTA_GTF_BAM_RPBP:RPBP_SELECTFINALPREDICTIONSET' {
+        ext.args   = { "--select-longest-by-stop --select-best-overlapping ${params.extra_rpbp_predictorfs_args ?: ''}".trim() }
+        publishDir = [
+            path: { "${params.outdir}/orf_predictions/rpbp" },
+            mode: params.publish_dir_mode,
+            pattern: "*.predicted-orfs.{bed.gz,dna.fa,protein.fa}"
+        ]
+    }
+
 }

docs/output.md

@@ -37,6 +37,8 @@ The pipeline is built using [Nextflow](https://www.nextflow.io/) and processes d
     - [Ribo-TISH predict](#ribo-tish-predict)
     - [Ribotricer detect-orfs](#ribotricer-detect-orfs)
     - [RiboCode](#ribocode)
+    - [Rp-Bp](#rp-bp)
+    - [PRICE](#price)
   - [P-site identification](#p-site-identification)
     - [riboWaltz](#ribowaltz)
     - [plastid](#plastid)
@@ -441,6 +443,20 @@ The `-f0_percent`, `-pv1`, and `-pv2` parameters belong to the **metaplots** ste
 
 If RiboCode is not needed for your analysis, you can skip it entirely with `--skip_ribocode`.
 
+### Rp-Bp
+
+<details markdown="1">
+<summary>Output files</summary>
+
+- `orf_predictions/rpbp/`
+  - `*.predicted-orfs.bed.gz`: per-sample predicted-ORF BED with Bayes factor scores (column 5) after the final-prediction-set filter (`--select-longest-by-stop --select-best-overlapping`).
+  - `*.predicted-orfs.dna.fa`: per-sample predicted-ORF nucleotide FASTA matching the BED.
+  - `*.predicted-orfs.protein.fa`: per-sample predicted-ORF protein FASTA matching the BED.
+
+</details>
+
+Produced only when `--run_rpbp true` is set. Rp-Bp's Bayesian fit is slow (~20-24h per replicate at genome-wide scale); see [Rp-Bp in usage.md](usage.md#rp-bp-opt-in-overnight). Rp-Bp's Bayes factor is stable across replicates and is retained in the cross-caller rank-aggregation set. When `--extended_orf_analysis true` is set, Rp-Bp consumes the hybrid GTF and so reports novel intergenic ORFs alongside annotated ones.
+
 ### PRICE
 
 <details markdown="1">

docs/usage.md

@@ -334,6 +334,22 @@ By default the pipeline calls ORFs with two tools, Ribo-TISH `predict` and RiboC
 
 Ribotricer is available as a third caller but is off by default. Enable it with `--run_ribotricer true` for broader recall, after which an ORF is agreed on a majority vote (2 of 3). It is opt-in because its ORF-score column is unstable across biological replicates even though its binary call set is reproducible. When enabled, its binary calls count toward agreement but its score is excluded from cross-caller rank aggregation, and the pipeline warns at runtime.
 
+### Rp-Bp (opt-in, overnight)
+
+[Rp-Bp](https://github.com/dieterich-lab/rp-bp) (Malone et al., 2017) is a Bayesian-strict ORF caller that complements RiboCode's permissive canonical-CDS calls. It is the recommended second caller when statistical rigour matters more than turnaround time. Activate with `--run_rpbp true`.
+
+> :warning: **Runtime cost.** Rp-Bp's Bayesian MCMC fit dominates wall-clock and takes roughly **20-24 hours per replicate** at genome-wide scale. The pipeline emits a runtime warning when `--run_rpbp` is set. Plan compute time, queue limits and instance lifetimes accordingly.
+
+Rp-Bp's score column (Bayes factor) is stable and is retained in the cross-caller rank-aggregation set alongside RiboCode and Ribo-TISH; Ribotricer's score column is excluded due to known instability but Rp-Bp's is not.
+
+Rp-Bp runs through the upstream `nf-core/rpbp/*` modules driven by the `FASTA_GTF_BAM_RPBP` nf-core subworkflow, which orchestrates `prepare-rpbp-genome`, `extract-metagene-profiles`, `estimate-metagene-profile-bayes-factors`, `select-periodic-offsets`, `get-periodic-lengths-offsets`, `extract-orf-profiles`, `estimate-orf-bayes-factors` and `select-final-prediction-set` from your `--fasta` / `--gtf` inputs without you having to author a YAML config. Tool CLI overrides are exposed via `--extra_rpbp_preparegenome_args` and `--extra_rpbp_predictorfs_args`.
+
+Per-sample final-prediction outputs - filtered BED of predicted ORFs (with Bayes factor in column 5), plus matched nucleotide and protein FASTAs - are published under `<outdir>/orf_predictions/rpbp/`.
+
+**Annotation.** Rp-Bp is given the full multi-isoform `--gtf` annotation, not the one-transcript-per-gene canonical backbone that the pipeline uses elsewhere to disambiguate P-site quantification. Rp-Bp enumerates candidate ORFs across every transcript isoform (deduplicating identical ORFs by genomic coordinate) and then resolves redundant and overlapping ORFs itself - the longest ORF per stop codon, then the highest Bayes factor among overlaps. Collapsing the annotation to one isoform per gene would silently remove ORFs that exist only on non-canonical isoforms (alternative-5'UTR uORFs, isoform-specific N-terminal extensions or truncations, retained-intron and alternative-exon ORFs) and bias the reported ORF types toward canonical CDS, with no compensating benefit; PRICE is handled the same way and for the same reason ([Malone et al., 2017](https://academic.oup.com/nar/article/45/6/2960/2953491)). Under `--extended_orf_analysis true` Rp-Bp instead receives the hybrid GTF, so novel transcripts are within discovery scope in the same way as Ribo-TISH `predict` and Ribotricer.
+
+> :information_source: **STAR alignment params vs upstream rpbp.** rpbp's own pipeline runs STAR with Ribo-seq-tuned settings (`outFilterMismatchNmax 1`, `outFilterMismatchNoverLmax 0.04`, `outFilterType BySJout`, `sjdbOverhang 33`, `winAnchorMultimapNmax 100`, `seedSearchStartLmaxOverLread 0.5`). We use the pipeline's standard STAR alignment (shared with the RNA-seq side of paired runs), which is more permissive. Practical impact: rpbp processes whatever alignments it gets, but periodicity / Bayes-factor distributions will differ from a standalone rpbp run on the same FASTQs. If you need bit-identical-to-standalone-rpbp output, override with `--extra_star_align_args '--outFilterMismatchNmax 1 --outFilterMismatchNoverLmax 0.04 --outFilterType BySJout --winAnchorMultimapNmax 100 --seedSearchStartLmaxOverLread 0.5'`. Note that `sjdbOverhang` is baked into the STAR index and cannot be changed post-hoc - it would require regenerating the index with `--sjdbOverhang 33`, and that change would only be appropriate for a Ribo-seq-only run (RNA-seq reads are too long for that setting). Tracked for future work: [#173](https://github.com/nf-core/riboseq/issues/173).
+
 ### PRICE (opt-in)
 
 [PRICE](https://github.com/erhard-lab/gedi/wiki/Price) (Erhard et al., 2018) is a Bayesian ORF caller distributed as part of the [Gedi](https://github.com/erhard-lab/gedi) Java framework. Unlike the per-sample callers, PRICE estimates a shared codon-position model across the riboseq cohort by EM and is invoked one-shot rather than per-sample. Activate with `--run_price true`.
@@ -342,6 +358,8 @@ Ribotricer is available as a third caller but is off by default. Enable it with
 
 The pipeline builds a binary `.oml` genome index via `gedi -e IndexGenome` once per run, then calls PRICE once across the cohort with the index plus the riboseq BAMs. PRICE's primary output is `${prefix}.orfs.tsv`, a table of all called ORFs with start-codon score, range score, p-value (uncorrected) and per-condition / total read counts. Tool CLI arguments can be appended via `--extra_price_indexgenome_args` and `--extra_price_price_args`.
 
+**Annotation.** Like Rp-Bp, PRICE is given the full multi-isoform `--gtf` annotation rather than the one-transcript-per-gene canonical backbone: it resolves overlapping ORFs and rescues multimappers with its own EM, so restricting it to a single isoform per gene would only narrow ORF discovery and bias ORF-type classification toward canonical CDS.
+
 When `--extended_orf_analysis true` is set, PRICE's IndexGenome receives the hybrid GTF so ORFs on novel intergenic transcripts are within its discovery scope.
 
 PRICE's CLI banner reports `Price version 1.0.4` while the Bioconda package is `gedi 1.0.6a` (Price is one tool inside the Gedi umbrella). The pipeline captures the package version via `gedi -e Version` for `versions.yml`.

modules.json

@@ -200,6 +200,46 @@
                         "git_sha": "b59f74e059a49fce82f19fbf684e2876da85ee39",
                         "installed_by": ["modules"]
                     },
+                    "rpbp/estimatemetagenebayesfactors": {
+                        "branch": "master",
+                        "git_sha": "617c552da369f63371648679983483736e52f3b8",
+                        "installed_by": ["fasta_gtf_bam_rpbp", "modules"]
+                    },
+                    "rpbp/estimateorfbayesfactors": {
+                        "branch": "master",
+                        "git_sha": "bbd642353cae3464d405ab6ae7366532648164e7",
+                        "installed_by": ["fasta_gtf_bam_rpbp", "modules"]
+                    },
+                    "rpbp/extractmetageneprofiles": {
+                        "branch": "master",
+                        "git_sha": "df05573454925ff87ff6ea6c4afaef70c64b7248",
+                        "installed_by": ["fasta_gtf_bam_rpbp", "modules"]
+                    },
+                    "rpbp/extractorfprofiles": {
+                        "branch": "master",
+                        "git_sha": "ad585e77451e4638e876bcf895185f0ac7f85fae",
+                        "installed_by": ["fasta_gtf_bam_rpbp", "modules"]
+                    },
+                    "rpbp/getperiodiclengthsoffsets": {
+                        "branch": "master",
+                        "git_sha": "caba7d6afa153b377581449609109fcd43775cab",
+                        "installed_by": ["fasta_gtf_bam_rpbp", "modules"]
+                    },
+                    "rpbp/preparegenome": {
+                        "branch": "master",
+                        "git_sha": "50b83e9601f411ef670b67b205db78e707a88f01",
+                        "installed_by": ["fasta_gtf_bam_rpbp", "modules"]
+                    },
+                    "rpbp/selectfinalpredictionset": {
+                        "branch": "master",
+                        "git_sha": "c2b3bb2252858a79872a27c7918cdb4b2a43f778",
+                        "installed_by": ["fasta_gtf_bam_rpbp", "modules"]
+                    },
+                    "rpbp/selectperiodicoffsets": {
+                        "branch": "master",
+                        "git_sha": "2f2f5662c0719221637811c6a6e443ec3414e0df",
+                        "installed_by": ["fasta_gtf_bam_rpbp", "modules"]
+                    },
                     "rsem/preparereference": {
                         "branch": "master",
                         "git_sha": "004e773fc35ebd24063ca4cbef057c94a24208aa",
@@ -377,6 +417,11 @@
                         "git_sha": "2fc6aef2691483864904e31973ccafd2ed68fd56",
                         "installed_by": ["subworkflows"]
                     },
+                    "fasta_gtf_bam_rpbp": {
+                        "branch": "master",
+                        "git_sha": "2111854cad9111c2b8057a6e050e0656d41acf32",
+                        "installed_by": ["subworkflows"]
+                    },
                     "fastq_align_star": {
                         "branch": "master",
                         "git_sha": "cebe21bbd158c15c8fab172e37cfe97a239f4b77",