clystere

clystere is a Nextflow pipeline for automated biosynthetic gene cluster (BGC) discovery and comparative analysis. It runs antiSMASH, GECCO, and deepBGC by default across a collection of genomes, unifies predictions with comBGC, and optionally groups resulting BGCs into gene cluster families (GCFs) with BiG-SCAPE or BiG-SLiCE.

---

Features

• Parallel antiSMASH + GECCO + deepBGC annotation across any number of genome assemblies or GenBank files
• comBGC-based unification of overlapping predictions from all three tools before clustering
• Per-region tabulation and per-genome BGC count summary
• Optional BiG-SCAPE or BiG-SLiCE clustering (mutually exclusive)
• Optional automatic bigscape dereplicate step before BiG-SCAPE clustering

---

Requirements

• Nextflow ≥ 23.04.0
• One of: Docker, Singularity, Podman, or Conda

---

Quick start

Run with Docker on the bundled example data

nextflow run exterex/clystere \
    --input assets/samplesheet.csv \
    --outdir results \
    -profile docker

---

Inputs

Samplesheet

A comma-separated file with the following columns:

Column	Required	Description
sample	Yes	Unique sample identifier
genome	Yes	Path to a genome file (GenBank, EMBL, or FASTA)
annotation	No	Path to a GFF3 annotation file (suppresses gene-finding)

sample,genome,annotation
strain_A,data/strain_A.gbff.gz,
strain_B,data/strain_B.fna,,
strain_C,data/strain_C.fna.gz,data/strain_C.gff3

antiSMASH database

A pre-built antiSMASH database directory. If --antismash_db points to a missing or empty directory the pipeline will download the database there automatically. The database is also resolvable via antismash-download-databases.

---

Parameters

A full parameter reference is available in nextflow_schema.json. Key parameters are summarised below.

Input / output

Parameter	Default	Description
--input	—	Path to samplesheet CSV (required)
--outdir	results	Directory for all pipeline outputs
--antismash_db	—	Path to antiSMASH database directory

antiSMASH

Parameter	Default	Description
--antismash_taxon	bacteria	Taxonomic scope (bacteria or fungi)
--antismash_minimal	true	Run in minimal mode; enable modules individually
--antismash_cb_knownclusters	false	Run KnownClusterBlast; adds similarity columns to all_regions.tsv
--antismash_genefinding_tool	prodigal	Gene caller when no annotation is supplied
--antismash_minlength	1000	Minimum sequence length (bp)
--antismash_accept_failure	false	Continue if antiSMASH fails for a sample
--antismash_extra_args	""	Arbitrary additional flags passed to antiSMASH

GECCO

Parameter	Default	Description
--gecco_run	true	Run GECCO BGC prediction
--gecco_extra_args	""	Additional arguments passed to GECCO

deepBGC

Parameter	Default	Description
--deepbgc_run	true	Run deepBGC prediction
--deepbgc_data_dir	—	Path to deepBGC model/Pfam downloads (auto-downloaded if absent)
--deepbgc_extra_args	""	Additional arguments passed to deepBGC

comBGC unification

Parameter	Default	Description
--combgc_min_length	3000	Minimum BGC length retained by comBGC
--combgc_contig_edge	2	Exclude BGCs close to contig edges in comBGC

BiG-SCAPE

BiG-SCAPE and BiG-SLiCE in clystere run on unified comBGC-filtered regions and require --gecco_run true and --deepbgc_run true.

Parameter	Default	Description
--bigscape_run	false	Enable BiG-SCAPE GCF clustering
--bigscape_dereplicate	true	Run bigscape dereplicate before clustering
--bigscape_dereplicate_cutoff	0.8	Similarity cutoff for dereplication
--bigscape_gcf_cutoffs	0.3 0.5 0.7	Space-separated list of distance cutoffs
--bigscape_mix	true	Combine all BGC classes into one network
--bigscape_include_singletons	true	Include singletons in the output

BiG-SLiCE

Parameter	Default	Description
--bigslice_run	false	Enable BiG-SLiCE clustering (mutually exclusive with --bigscape_run)
--bigslice_extra_args	""	Additional arguments passed to BiG-SLiCE
--bigslice_zip_output	false	Compress BiG-SLiCE output directory

Tabulation

Parameter	Default	Description
--run_tabulation	true	Generate all_regions.tsv and region_counts.tsv
--count_per_contig	false	Report counts per contig rather than per assembly
--split_hybrids	false	Count each product type in hybrid BGCs separately

---

Outputs

results/
├── antismash/
│   └── <sample>/          # Full antiSMASH output per genome
├── gecco/
│   └── <sample>/          # GECCO outputs per genome (+ BiG-SLiCE-compatible regions)
├── deepbgc/
│   └── <sample>/          # deepBGC outputs per genome (+ converted region GBKs)
├── combgc/
│   └── <sample>/
│       ├── combgc_summary.tsv
│       └── combined_regions/   # Unified representative region GBKs used for clustering
├── bigscape/              # BiG-SCAPE output (when --bigscape_run)
├── bigslice/              # BiG-SLiCE output (when --bigslice_run)
├── summary/
│   ├── all_regions.tsv    # One row per BGC region across all samples
│   └── region_counts.tsv  # BGC type counts per genome (or per contig)
└── pipeline_info/         # Execution timeline, report, trace, and DAG

all_regions.tsv

Column	Description
file	Source antiSMASH run (genome stem)
record_id	Sequence/contig identifier
region	Region number within the record
start / end	Genomic coordinates (bp)
contig_edge	Whether the region extends to a contig boundary
product	BGC product class(es)
knownclusterblast_hit	Top MIBiG hit description (only when --antismash_cb_knownclusters)
knownclusterblast_accession	MIBiG accession (only when --antismash_cb_knownclusters)
knownclusterblast_similarity	Similarity category: low, medium, or high (only when --antismash_cb_knownclusters)
record_desc	Sequence description from the source file

region_counts.tsv

One row per genome (or per contig with --count_per_contig) with integer counts for each BGC product class detected, plus total_count and description columns.

---

Profiles

# Docker (default for local runs)
-profile docker

# Singularity (recommended for HPC)
-profile singularity

# Apptainer (recommended for HPC)
-profile apptainer

# Conda
-profile conda

# SLURM cluster — sets executor + raises resource ceilings
-profile singularity,slurm

# Generic HPC — raises resource ceilings without binding to a scheduler
-profile singularity,hpc

---

Citations

Please cite the pipeline and its dependencies. See CITATIONS.md for full references.

---

Licence

MIT