nextflow.config

// ============================================================================
// phinder — phage finder for metagenomic assemblies
// ============================================================================

manifest {
    name            = 'phinder'
    description     = 'Phage finder for metagenomic assemblies (geNomad + CheckV + Pharokka + PhaBOX)'
    author          = 'Andrew Budge'
    version         = '0.2.0'
    defaultBranch   = 'main'
    nextflowVersion = '>=23.04'
}

params {
    // --- Input / output -----------------------------------------------------
    input  = null            // required: combined contig FASTA (.fa/.fa.gz)
    outdir = 'results'

    // --- Resources ----------------------------------------------------------
    // Two ways to size phinder:
    //   • Convenience (default): use a FRACTION of the detected machine — half
    //     the cores and RAM. Polite on shared boxes, safe on small ones. Tune
    //     with --cpu_fraction / --mem_fraction (1.0 = the whole machine).
    //   • Ultimate control: set an ABSOLUTE budget with --max_cpus (cores) and
    //     --max_memory (e.g. '64.GB'). When set, these override the fractions.
    // If machine auto-detection is wrong, override --avail_cpus / --avail_mem.
    // See the README "Performance & resources" section.
    cpu_fraction = 0.5
    mem_fraction = 0.5
    avail_cpus   = Runtime.runtime.availableProcessors()
    avail_mem    = java.lang.management.ManagementFactory.operatingSystemMXBean.totalMemorySize
    max_cpus     = null      // exact core budget; overrides cpu_fraction when set
    max_memory   = null      // exact memory budget, e.g. '64.GB'; overrides mem_fraction

    // Effective budget the heavy tier gets: absolute override if given, else a
    // fraction of the detected machine. (mem_budget is in bytes.)
    cpu_budget   = params.max_cpus   ? (params.max_cpus as int)                                  : Math.max(1, (params.avail_cpus * params.cpu_fraction) as int)
    mem_budget   = params.max_memory ? (params.max_memory as nextflow.util.MemoryUnit).toBytes() : (params.avail_mem * params.mem_fraction) as long

    // --- Databases ----------------------------------------------------------
    genomad_db  = null    // required
    checkv_db   = null    // required
    pharokka_db = null    // required
    phabox_db   = null    // required only if phabox2_env is set

    // --- Conda env overrides (optional) -------------------------------------
    // Point at existing named envs instead of letting Nextflow build new ones.
    // Useful when tools have complex or non-standard installations.
    genomad_env  = null   // e.g. ~/miniconda3/envs/genomad
    checkv_env   = null   // e.g. ~/miniconda3/envs/checkv
    pharokka_env = null   // e.g. ~/miniconda3/envs/pharokka
    phabox2_env  = null   // e.g. ~/miniconda3/envs/phabox2; if unset, PhaBOX steps are skipped

    // --- geNomad ------------------------------------------------------------
    genomad_splits = 20

    // --- geNomad filter thresholds ------------------------------------------
    // DTR topology is always kept regardless of virus_score (hallmark of
    // complete linear phage genome). Provirus topology is kept only above
    // this score.
    min_provirus_score = 0.9

    // --- CheckV filter thresholds -------------------------------------------
    // Keep contigs with CheckV quality in this set, OR with DTR topology
    // (DTR contigs under-score on completeness because they lack host flanking).
    // Coverage below min_coverage is flagged but NOT dropped — annotation
    // stages get the final call.
    checkv_quality_keep = 'High-quality,Complete'
    min_coverage        = 3

    // --- Pharokka -----------------------------------------------------------
    pharokka_gene_predictor = 'prodigal-gv'

    // --- PhaBOX -------------------------------------------------------------
    // Skip Phamer (--skip Y) because sequences are already classified viral
    // upstream by geNomad + CheckV.
    phabox_skip_phamer  = true
    phabox_votu_mode    = 'AAI'     // protein-level; phages are too divergent for ANI
    phabox_tree_markers = 'terl,portal'
}

profiles {
    conda {
        conda.enabled = true
    }
    mamba {
        conda.enabled    = true
        conda.useMamba   = true
    }
    docker {
        docker.enabled = true
        // Run as the host user so output files aren't owned by root.
        docker.runOptions = '-u $(id -u):$(id -g)'
    }
    singularity {
        singularity.enabled    = true
        singularity.autoMounts = true
    }
    apptainer {
        apptainer.enabled    = true
        apptainer.autoMounts = true
    }
    slurm {
        process.executor = 'slurm'
    }
    test {
        // Minimal inputs to validate workflow wiring with `-stub-run`.
        // The bundled FASTA is real; the db paths are placeholders that the
        // stub blocks never read (no databases required).
        params.input       = "${projectDir}/test/test_contigs.fasta"
        params.genomad_db  = "${projectDir}/test"
        params.checkv_db   = "${projectDir}/test"
        params.pharokka_db = "${projectDir}/test"
    }
}

process {
    // Budget = params.cpu_budget / params.mem_budget (absolute --max_* if set,
    // else a fraction of the detected machine). No task may exceed it
    // (resourceLimits); the heavy tier uses the whole budget, lighter tiers a
    // share. With the default (half the machine) requests stay well under what's
    // available, so a run never fails with "requirement exceeds available".
    resourceLimits = [
        cpus  : params.cpu_budget,
        memory: params.mem_budget as long
    ]

    withLabel: process_low {       // seqkit / R / sed glue — tiny
        cpus   = 1
        memory = (params.mem_budget / 4) as long
    }
    withLabel: process_medium {    // pharokka, phabox
        cpus   = Math.max(1, (params.cpu_budget / 2) as int)
        memory = (params.mem_budget / 2) as long
    }
    withLabel: process_high {      // genomad, checkv — the heavy lifters
        cpus   = params.cpu_budget
        memory = params.mem_budget as long
    }
}