From 3ca17f1f829a868f666c614f52fd9a5f5ef595cb Mon Sep 17 00:00:00 2001
From: alejandrogzi <jose.gonzalesdezavala1@unmsm.edu.pe>
Date: Wed, 30 Sep 2026 13:25:41 +0200
Subject: [PATCH] feat(oriongeno): add Nextflow and WDL annotation modules

Wrap ghcr.io/hillerlab/oriongeno for single-GPU prediction and
multi-GPU annotation. Lineage selects the checkpoint; species name
does not. Predict writes GTF by default and GFF3 on request.
---
 assets/catalog/catalog.md                  |   2 +
 modules/nextflow/oriongeno/multi/main.nf   | 114 +++++++++++++++++++
 modules/nextflow/oriongeno/predict/main.nf | 105 ++++++++++++++++++
 modules/wdl/oriongeno/multi/main.wdl       | 119 ++++++++++++++++++++
 modules/wdl/oriongeno/predict/main.wdl     | 122 +++++++++++++++++++++
 5 files changed, 462 insertions(+)
 create mode 100644 modules/nextflow/oriongeno/multi/main.nf
 create mode 100644 modules/nextflow/oriongeno/predict/main.nf
 create mode 100644 modules/wdl/oriongeno/multi/main.wdl
 create mode 100644 modules/wdl/oriongeno/predict/main.wdl

diff --git a/assets/catalog/catalog.md b/assets/catalog/catalog.md
index 09dfd85..19ab2b9 100644
--- a/assets/catalog/catalog.md
+++ b/assets/catalog/catalog.md
@@ -120,6 +120,8 @@
 - [tiberius/merge](../../modules/nextflow/tiberius/merge/main.nf) — Merge chunk GTFs with merge_annotations.py, restore FASTA order, emit GTF/GFF3 and optional protein/CDS FASTAs.
 - [subworkflows/tiberius](../../subworkflows/tiberius/main.nf) — Scatter (none / chromosome / weighted), predict, and merge.
 - [workflows/tiberius](../../workflows/tiberius/main.nf) — Standalone Tiberius pipeline.
+- [oriongeno/predict](../../modules/nextflow/oriongeno/predict/main.nf) — OrionGeno ab initio gene annotation on one GPU (`process_gpu`). No CPU fallback. Default checkpoint: mammals. GTF by default, GFF3 via `meta.oriongeno_format`. Bundled OrionGeno is non-commercial.
+- [oriongeno/multi](../../modules/nextflow/oriongeno/multi/main.nf) — OrionGeno annotation across the GPUs assigned to the task (`oriongeno multi`). Writes GTF. `process_gpu` requests one GPU; raise `accelerator` on `ORIONGENO_MULTI` to use more.
 
 ---
 
diff --git a/modules/nextflow/oriongeno/multi/main.nf b/modules/nextflow/oriongeno/multi/main.nf
new file mode 100644
index 0000000..a3dd0df
--- /dev/null
+++ b/modules/nextflow/oriongeno/multi/main.nf
@@ -0,0 +1,114 @@
+/*
+Copyright (c) 2026 The Hiller Lab at the Senckenberg Gessellschaft für Naturforschung
+Distributed under the terms of the Apache License, Version 2.0.
+*/
+
+/*
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+    ORIONGENO_MULTI — Ab initio OrionGeno annotation across one or more GPUs.
+
+    Thin wrapper around `oriongeno multi`. `--devices` is the GPU list
+    Nextflow assigned in CUDA_VISIBLE_DEVICES, or 0 when that variable is
+    unset. `label 'process_gpu'` requests one GPU under `-profile gpu`.
+    To use several GPUs, set accelerator for this process in the config;
+    that selector overrides the label:
+
+        process {
+            withName: 'ORIONGENO_MULTI' {
+                accelerator = 4
+            }
+        }
+
+    `oriongeno multi` writes GTF. meta.oriongeno_format other than gtf is
+    an error. Lineage, species, repeat output, and extra args match
+    ORIONGENO_PREDICT. Inference requires an NVIDIA GPU; there is no CPU
+    fallback. Bundled OrionGeno is non-commercial.
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+*/
+
+process ORIONGENO_MULTI {
+    tag "$meta.id"
+    label 'process_high'
+    label 'process_gpu'
+
+    container "${ workflow.containerEngine == 'singularity' && !task.ext.singularity_pull_docker_container ?
+        '' :
+        'ghcr.io/hillerlab/oriongeno:latest' }"
+
+    input:
+    tuple val(meta), path(genome)
+    val lineage
+
+    output:
+    tuple val(meta), path("annotation/*"), emit: annotation
+    tuple val(meta), path("repeats/*"),    emit: repeats, optional: true
+    path "versions.yml",                   emit: versions
+
+    when:
+    task.ext.when == null || task.ext.when
+
+    script:
+    def args = [task.ext.args ?: '', meta.oriongeno_args ?: ''].findAll { it }.join(' ')
+    def prefix = task.ext.prefix ?: "${meta.id}"
+    def known = ['mammals', 'birds', 'fish', 'other_vertebrates', 'arthropods', 'other_invertebrates', 'plants', 'fungi']
+    def lineage_name = lineage.toString()
+    def checkpoint = (task.ext.checkpoint ?: meta.checkpoint ?: '').toString()
+    if (!checkpoint) {
+        if (!known.contains(lineage_name)) {
+            error "Unknown OrionGeno lineage '${lineage_name}'. Expected one of: ${known.join(', ')}. Or set task.ext.checkpoint / meta.checkpoint to a checkpoint directory."
+        }
+        checkpoint = "/opt/oriongeno/checkpoints/oriongeno_${lineage_name}"
+    }
+    def format = (meta.oriongeno_format ?: 'gtf').toString().toLowerCase()
+    if (format != 'gtf') {
+        error "ORIONGENO_MULTI writes GTF. oriongeno multi does not emit GFF."
+    }
+    def species = (task.ext.species_name ?: meta.species_name ?: meta.species ?: '').toString()
+    def species_arg = species ? "--species-name \"${species}\"" : ''
+    def emit_repeat = meta.oriongeno_output_repeat == true || meta.oriongeno_output_repeat == 'true'
+    """
+    set -euo pipefail
+    mkdir -p annotation
+
+    devices="\${CUDA_VISIBLE_DEVICES:-0}"
+    if [ -z "\$devices" ]; then
+        devices=0
+    fi
+
+    oriongeno multi \\
+        --genome "${genome}" \\
+        --output "annotation/${prefix}.gtf" \\
+        --checkpoint "${checkpoint}" \\
+        --devices "\$devices" \\
+        --work-dir oriongeno-work \\
+        --output-gene true \\
+        --output-repeat ${emit_repeat ? 'true' : 'false'} \\
+        ${species_arg} \\
+        ${args}
+
+    if [ -f "annotation/${prefix}.repeat.gtf" ]; then
+        mkdir -p repeats
+        mv "annotation/${prefix}.repeat.gtf" "repeats/${prefix}.repeat.gtf"
+    fi
+
+    cat <<-END_VERSIONS > versions.yml
+    "${task.process}":
+        oriongeno: \$( oriongeno --version 2>/dev/null | sed 's/^oriongeno //' || echo "2026.7.10" )
+    END_VERSIONS
+    """
+
+    stub:
+    def prefix = task.ext.prefix ?: "${meta.id}"
+    def emit_repeat = meta.oriongeno_output_repeat == true || meta.oriongeno_output_repeat == 'true'
+    """
+    set -euo pipefail
+    mkdir -p annotation
+    touch "annotation/${prefix}.gtf"
+    ${emit_repeat ? "mkdir -p repeats && touch \"repeats/${prefix}.repeat.gtf\"" : ''}
+
+    cat <<-END_VERSIONS > versions.yml
+    "${task.process}":
+        oriongeno: \$( oriongeno --version 2>/dev/null | sed 's/^oriongeno //' || echo "2026.7.10" )
+    END_VERSIONS
+    """
+}
diff --git a/modules/nextflow/oriongeno/predict/main.nf b/modules/nextflow/oriongeno/predict/main.nf
new file mode 100644
index 0000000..d28f281
--- /dev/null
+++ b/modules/nextflow/oriongeno/predict/main.nf
@@ -0,0 +1,105 @@
+/*
+Copyright (c) 2026 The Hiller Lab at the Senckenberg Gessellschaft für Naturforschung
+Distributed under the terms of the Apache License, Version 2.0.
+*/
+
+/*
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+    ORIONGENO_PREDICT — Ab initio OrionGeno annotation on one GPU.
+
+    Thin wrapper around `oriongeno`. Inference requires an NVIDIA GPU
+    (compute capability >= 7.0, driver >= 525.60.13). There is no CPU fallback.
+    `lineage` selects a checkpoint baked into the image. It does not select
+    the species. The published image contains mammals unless it was built
+    with ORIONGENO_MODELS. A directory in task.ext.checkpoint or
+    meta.checkpoint overrides the lineage. Species conditioning is
+    meta.species_name (or meta.species, or task.ext.species_name).
+    meta.oriongeno_format is gtf (default) or gff3. meta.oriongeno_output_repeat
+    also writes the repeat annotation. Extra flags go in task.ext.args or
+    meta.oriongeno_args and are not given biological defaults here.
+    Bundled OrionGeno is non-commercial.
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+*/
+
+process ORIONGENO_PREDICT {
+    tag "$meta.id"
+    label 'process_high'
+    label 'process_gpu'
+
+    container "${ workflow.containerEngine == 'singularity' && !task.ext.singularity_pull_docker_container ?
+        '' :
+        'ghcr.io/hillerlab/oriongeno:latest' }"
+
+    input:
+    tuple val(meta), path(genome)
+    val lineage
+
+    output:
+    tuple val(meta), path("annotation/*"), emit: annotation
+    tuple val(meta), path("repeats/*"),    emit: repeats, optional: true
+    path "versions.yml",                   emit: versions
+
+    when:
+    task.ext.when == null || task.ext.when
+
+    script:
+    def args = [task.ext.args ?: '', meta.oriongeno_args ?: ''].findAll { it }.join(' ')
+    def prefix = task.ext.prefix ?: "${meta.id}"
+    def known = ['mammals', 'birds', 'fish', 'other_vertebrates', 'arthropods', 'other_invertebrates', 'plants', 'fungi']
+    def lineage_name = lineage.toString()
+    def checkpoint = (task.ext.checkpoint ?: meta.checkpoint ?: '').toString()
+    if (!checkpoint) {
+        if (!known.contains(lineage_name)) {
+            error "Unknown OrionGeno lineage '${lineage_name}'. Expected one of: ${known.join(', ')}. Or set task.ext.checkpoint / meta.checkpoint to a checkpoint directory."
+        }
+        checkpoint = "/opt/oriongeno/checkpoints/oriongeno_${lineage_name}"
+    }
+    def format = (meta.oriongeno_format ?: 'gtf').toString().toLowerCase()
+    if (!(format in ['gtf', 'gff', 'gff3'])) {
+        error "Unknown OrionGeno format '${format}'. Use gtf or gff3."
+    }
+    def ext = (format == 'gff' || format == 'gff3') ? 'gff3' : 'gtf'
+    def species = (task.ext.species_name ?: meta.species_name ?: meta.species ?: '').toString()
+    def species_arg = species ? "--species-name \"${species}\"" : ''
+    def emit_repeat = meta.oriongeno_output_repeat == true || meta.oriongeno_output_repeat == 'true'
+    """
+    set -euo pipefail
+    mkdir -p annotation
+
+    oriongeno \\
+        --genome "${genome}" \\
+        --output "annotation/${prefix}.${ext}" \\
+        --checkpoint "${checkpoint}" \\
+        --output-gene true \\
+        --output-repeat ${emit_repeat ? 'true' : 'false'} \\
+        ${species_arg} \\
+        ${args}
+
+    if [ -f "annotation/${prefix}.repeat.${ext}" ]; then
+        mkdir -p repeats
+        mv "annotation/${prefix}.repeat.${ext}" "repeats/${prefix}.repeat.${ext}"
+    fi
+
+    cat <<-END_VERSIONS > versions.yml
+    "${task.process}":
+        oriongeno: \$( oriongeno --version 2>/dev/null | sed 's/^oriongeno //' || echo "2026.7.10" )
+    END_VERSIONS
+    """
+
+    stub:
+    def prefix = task.ext.prefix ?: "${meta.id}"
+    def format = (meta.oriongeno_format ?: 'gtf').toString().toLowerCase()
+    def ext = (format == 'gff' || format == 'gff3') ? 'gff3' : 'gtf'
+    def emit_repeat = meta.oriongeno_output_repeat == true || meta.oriongeno_output_repeat == 'true'
+    """
+    set -euo pipefail
+    mkdir -p annotation
+    touch "annotation/${prefix}.${ext}"
+    ${emit_repeat ? "mkdir -p repeats && touch \"repeats/${prefix}.repeat.${ext}\"" : ''}
+
+    cat <<-END_VERSIONS > versions.yml
+    "${task.process}":
+        oriongeno: \$( oriongeno --version 2>/dev/null | sed 's/^oriongeno //' || echo "2026.7.10" )
+    END_VERSIONS
+    """
+}
diff --git a/modules/wdl/oriongeno/multi/main.wdl b/modules/wdl/oriongeno/multi/main.wdl
new file mode 100644
index 0000000..e29fa17
--- /dev/null
+++ b/modules/wdl/oriongeno/multi/main.wdl
@@ -0,0 +1,119 @@
+# Copyright (c) 2026 The Hiller Lab at the Senckenberg Gessellschaft für Naturforschung
+# Distributed under the terms of the Apache License, Version 2.0.
+
+# ORIONGENO_MULTI — Ab initio OrionGeno annotation across one or more GPUs.
+# Thin wrapper around `oriongeno multi`. devices is a comma-separated list of
+# GPU ids ("0,1,2,3"). The caller schedules those devices; this task writes GTF.
+# lineage, species_name, output_repeat, checkpoint, and extra_args match predict.
+# Inference requires an NVIDIA GPU. There is no CPU fallback.
+# Bundled OrionGeno is non-commercial.
+
+version 1.3
+
+task multi {
+  input {
+    File genome
+    String devices
+    String lineage = "mammals"
+    String species_name = ""
+    Boolean output_repeat = false
+    String checkpoint = ""
+    String extra_args = ""
+    String prefix = ""
+  }
+
+  String stem = sub(sub(basename(genome), "\\.gz$", ""), "\\.(fa|fasta|fna|fas)$", "")
+  String sample = if prefix != "" then prefix else stem
+  String annotation_path = "annotation/" + sample + ".gtf"
+  String repeat_path = "repeats/" + sample + ".repeat.gtf"
+
+  command <<<
+    set -euo pipefail
+
+    lineage="~{lineage}"
+    devices="~{devices}"
+    checkpoint="~{checkpoint}"
+    species="~{species_name}"
+
+    if [ -z "$devices" ]; then
+      echo "ORIONGENO_MULTI requires at least one GPU id in devices." >&2
+      exit 1
+    fi
+
+    if [ -n "$checkpoint" ]; then
+      ckpt="$checkpoint"
+    else
+      case "$lineage" in
+        mammals|birds|fish|other_vertebrates|arthropods|other_invertebrates|plants|fungi) ;;
+        *)
+          echo "Unknown OrionGeno lineage '$lineage'." >&2
+          exit 1
+          ;;
+      esac
+      ckpt="/opt/oriongeno/checkpoints/oriongeno_${lineage}"
+    fi
+
+    mkdir -p annotation
+    species_args=()
+    if [ -n "$species" ]; then
+      species_args+=(--species-name "$species")
+    fi
+
+    oriongeno multi \
+      --genome "~{genome}" \
+      --output "~{annotation_path}" \
+      --checkpoint "$ckpt" \
+      --devices "$devices" \
+      --work-dir oriongeno-work \
+      --output-gene true \
+      --output-repeat ~{if output_repeat then "true" else "false"} \
+      "${species_args[@]}" \
+      ~{extra_args}
+
+    gene="~{annotation_path}"
+    repeat_src="${gene%.*}.repeat.${gene##*.}"
+    if [ -f "$repeat_src" ]; then
+      mkdir -p repeats
+      mv "$repeat_src" "~{repeat_path}"
+    fi
+  >>>
+
+  output {
+    File annotation = annotation_path
+    File? repeats = repeat_path
+  }
+
+  requirements {
+    container: "ghcr.io/hillerlab/oriongeno:latest"
+  }
+}
+
+workflow run {
+  input {
+    File genome
+    String devices
+    String lineage = "mammals"
+    String species_name = ""
+    Boolean output_repeat = false
+    String checkpoint = ""
+    String extra_args = ""
+    String prefix = ""
+  }
+
+  call multi {
+    input:
+      genome = genome,
+      devices = devices,
+      lineage = lineage,
+      species_name = species_name,
+      output_repeat = output_repeat,
+      checkpoint = checkpoint,
+      extra_args = extra_args,
+      prefix = prefix
+  }
+
+  output {
+    File annotation = multi.annotation
+    File? repeats = multi.repeats
+  }
+}
diff --git a/modules/wdl/oriongeno/predict/main.wdl b/modules/wdl/oriongeno/predict/main.wdl
new file mode 100644
index 0000000..76bcbd3
--- /dev/null
+++ b/modules/wdl/oriongeno/predict/main.wdl
@@ -0,0 +1,122 @@
+# Copyright (c) 2026 The Hiller Lab at the Senckenberg Gessellschaft für Naturforschung
+# Distributed under the terms of the Apache License, Version 2.0.
+
+# ORIONGENO_PREDICT — Ab initio OrionGeno annotation on one GPU.
+# Inference requires an NVIDIA GPU (compute capability >= 7.0, driver >= 525.60.13).
+# There is no CPU fallback. lineage selects a checkpoint baked into the image
+# and does not select the species. The published image contains mammals unless
+# it was built with ORIONGENO_MODELS. checkpoint overrides lineage when set.
+# format is gtf (default) or gff3. output_repeat also writes the repeat annotation.
+# extra_args is appended as-is. Bundled OrionGeno is non-commercial.
+
+version 1.3
+
+task predict {
+  input {
+    File genome
+    String lineage = "mammals"
+    String species_name = ""
+    String format = "gtf"
+    Boolean output_repeat = false
+    String checkpoint = ""
+    String extra_args = ""
+    String prefix = ""
+  }
+
+  String stem = sub(sub(basename(genome), "\\.gz$", ""), "\\.(fa|fasta|fna|fas)$", "")
+  String sample = if prefix != "" then prefix else stem
+  String ext = if format == "gff" || format == "gff3" then "gff3" else "gtf"
+  String annotation_path = "annotation/" + sample + "." + ext
+  String repeat_path = "repeats/" + sample + ".repeat." + ext
+
+  command <<<
+    set -euo pipefail
+
+    lineage="~{lineage}"
+    format="~{format}"
+    checkpoint="~{checkpoint}"
+    species="~{species_name}"
+
+    case "$format" in
+      gtf|gff|gff3) ;;
+      *)
+        echo "Unknown OrionGeno format '$format'. Use gtf or gff3." >&2
+        exit 1
+        ;;
+    esac
+
+    if [ -n "$checkpoint" ]; then
+      ckpt="$checkpoint"
+    else
+      case "$lineage" in
+        mammals|birds|fish|other_vertebrates|arthropods|other_invertebrates|plants|fungi) ;;
+        *)
+          echo "Unknown OrionGeno lineage '$lineage'." >&2
+          exit 1
+          ;;
+      esac
+      ckpt="/opt/oriongeno/checkpoints/oriongeno_${lineage}"
+    fi
+
+    mkdir -p annotation
+    species_args=()
+    if [ -n "$species" ]; then
+      species_args+=(--species-name "$species")
+    fi
+
+    oriongeno \
+      --genome "~{genome}" \
+      --output "~{annotation_path}" \
+      --checkpoint "$ckpt" \
+      --output-gene true \
+      --output-repeat ~{if output_repeat then "true" else "false"} \
+      "${species_args[@]}" \
+      ~{extra_args}
+
+    gene="~{annotation_path}"
+    repeat_src="${gene%.*}.repeat.${gene##*.}"
+    if [ -f "$repeat_src" ]; then
+      mkdir -p repeats
+      mv "$repeat_src" "~{repeat_path}"
+    fi
+  >>>
+
+  output {
+    File annotation = annotation_path
+    File? repeats = repeat_path
+  }
+
+  requirements {
+    container: "ghcr.io/hillerlab/oriongeno:latest"
+  }
+}
+
+workflow run {
+  input {
+    File genome
+    String lineage = "mammals"
+    String species_name = ""
+    String format = "gtf"
+    Boolean output_repeat = false
+    String checkpoint = ""
+    String extra_args = ""
+    String prefix = ""
+  }
+
+  call predict {
+    input:
+      genome = genome,
+      lineage = lineage,
+      species_name = species_name,
+      format = format,
+      output_repeat = output_repeat,
+      checkpoint = checkpoint,
+      extra_args = extra_args,
+      prefix = prefix
+  }
+
+  output {
+    File annotation = predict.annotation
+    File? repeats = predict.repeats
+  }
+}