FunFlux

__________             _______________
___  ____/___  ___________  ____/__  /___  _____  __
__  /_   _  / / /_  __ \_  /_   __  /_  / / /_  |/_/
_  __/   / /_/ /_  / / /  __/   _  / / /_/ /__>  <
/_/      \__,_/ /_/ /_//_/      /_/  \__,_/ /_/|_|

FunFlux v2.1.0

Fungal genome assembly and annotation from Illumina short reads, ONT or PacBio HiFi long reads, a hybrid of short and long reads, or an assembly you already have.

The detailed map, step by step, is on the documentation home page.

📖 Documentation

Full documentation: iLivius.github.io/FunFlux

This README is deliberately short: it covers what FunFlux is and how to get a first run going. Everything else — choosing a mode, installation, the reference databases and licensed tools, every configuration key, the output files, running on a cluster, troubleshooting, and the validation on public genomes — lives on the documentation site.

Synopsis

FunFlux is a Snakemake workflow that takes a fungal isolate from raw reads — or from an assembly somebody else made — through assembly, contamination and mitochondrion screening, quality control, ITS taxonomy, repeat masking, gene prediction and functional annotation, with one MultiQC report per run. One key in the configuration, mode, decides which of four front ends runs:

               illumina        longread           hybrid                contigs
01.reads       PhiX removal    fastplong (ONT)    PhiX removal, fastp   -
               fastp           Filtlong           fastplong (ONT)
                               NanoPlot           Filtlong, NanoPlot
02.assembly    SPAdes          hifiasm or Flye    SPAdes guide,         contig filter
                                                  hifiasm or Flye
               ── contamination screen (BLAST+, BlobTools), mitochondrion screen * ──
               ── adapter screen of the draft contigs (BLAST+) ──
                               Medaka (ONT)       Medaka (ONT),
                                                  Polypolish
               ── 02.assembly/{sample}/contigs_final.fasta: the genome all later steps read ──
02.assembly    QUAST, BUSCO
03.taxonomy    ITS: 5.8S windows (nhmmer), ITSx, SINTAX against UNITE
04.annotation  funannotate clean and sort; tantan or RepeatModeler + RepeatMasker;
               Tiberius ‡ and funannotate predict (GeneMark-ES †); InterProScan,
               eggNOG-mapper, antiSMASH, SignalP 6 †; funannotate annotate
05.amr         resistance mutations: ChroQueTas against FungAMR †
06.report      MultiQC

* contigs mode: only with parameters.contigs.screen: true    † optional    ‡ on by default

FunFlux was developed for the genomic characterisation of fungal isolates. Its results describe the assembled genome, its predicted genes and their functional annotation, including the secondary-metabolite gene clusters found by antiSMASH. They are predictions from the genome sequence, to be weighed with curation and experimental evidence.

Each mode produces 02.assembly/{sample}/contigs_final.fasta, the genome that every later analysis reads. When the mitochondrion screen runs with its default settings, the confirmed mitochondrial contigs are split off into contigs_mitochondrial.fasta, and contigs_final.fasta holds the nuclear genome alone. The screen always runs in illumina, longread and hybrid, and in contigs mode only when it is switched on. Common processing improves consistency across datasets, while the read type, the sequencing depth and the fragmentation of the assembly still affect the results.

The resistance-mutation screen (05.amr, disabled by default) searches each genome with ChroQueTas for the point mutations that the FungAMR database lists with evidence for or against resistance to antifungal drugs and fungicides. It reports known mutations in the genes and species the database covers. It does not establish a resistance phenotype, and the absence of a call is not evidence of susceptibility. See Antifungal resistance mutations.

v2.0.0 brings together the earlier FunFlux 1.x, Funnotator and FunFluxL workflows as modes of one workflow. The earlier versions retain their DOIs and remain citable for work performed with them. A v1 configuration does not run in v2. If you are coming from one of them, start at Coming from v1.

FunFlux belongs to the BioFlux family of pipelines.

Quick start

# 1. Clone
git clone https://github.com/iLivius/FunFlux.git
cd FunFlux

# 2. Launcher environment: Snakemake 9.14.6 or a later 9.x release (FunFlux v2 was tested
#    with 9.14.6 and stops on anything older; 10 and later have not been tested)
conda create -n snakemake -c conda-forge -c bioconda "snakemake>=9.14.6,<10"
conda activate snakemake

# 3. Copy the template and edit the copy: mode, the input directory of that mode,
#    output_dir, the database paths, files.iprscan and files.annotation_params.
#    config/config_custom.yaml is git-ignored.
cp config/config.yaml config/config_custom.yaml

# 4. Dry run, then run. --cores takes the value of resources.threads (24 in the template).
snakemake --sdm conda --cores 24 --configfile config/config_custom.yaml -n
snakemake --sdm conda --cores 24 --configfile config/config_custom.yaml

Run both commands from the repository root: workflow/Snakefile is found there automatically. Besides Snakemake, three tools are not installed by conda: InterProScan, which every run needs, and the optional GeneMark-ES and SignalP 6, which need a licence from their authors. You obtain them yourself (Licensed and external tools). --sdm conda builds each rule's own conda environment on first use, so every other analysis tool is provisioned for you. The databases are listed below.

For a local run, --cores N sets the CPU budget and should take the value of resources.threads. Do not add --jobs/-j: for a local run it is another name for --cores, and giving both can let the jobs use more CPUs than the machine has. See Running FunFlux.

Always dry-run first. The configuration is checked while the plan is being built, so -n takes seconds and stops on a wrong mode, a missing input directory, a bad sample name or a missing annotation row before the first job starts.

Every sample needs a row in the tab-separated table named by files.annotation_params. config/annotation_parameters.tsv is an example:

Column	Content
1	sample name
2	species
3	a protein FASTA of related fungi (gene evidence)
4, optional	a stored AUGUSTUS model to use instead of training AUGUSTUS on the genome

Three resources must be on disk before the first run, and the two optional licensed tools if you use them:

Config key	What	Version	Size
directories.blast_db	NCBI core_nt with taxdb.btd/taxdb.bti and the taxdump files nodes.dmp/names.dmp in the same folder; not read in contigs mode without the screen	—	264 GB (build of 11 July 2026, index files included; the taxonomy files add 0.8 GB)
directories.eggnog_db	eggNOG-mapper data (eggnog.db, eggnog_proteins.dmnd)	eggNOG 5.0.2	48 GB
files.iprscan	InterProScan, the path of interproscan.sh	5.78-109.0	50 GB after setup
directories.genemark_dir	GeneMark-ES with its licence key in ~/.gm_key; optional	4	—
files.signalp6_archive	SignalP 6 archive from DTU; optional, switched on with parameters.signalp6.mode	6.0	—

The eggNOG version is fixed by eggnog-mapper 2.1.13, whose download_eggnog_data.py fetches eggNOG 5.0.2, so data made for eggnog-mapper 3 (eggNOG 7) does not work with it.

Everything else — the funannotate database snapshot (Zenodo record 23265189, checked against its SHA-256 before unpacking), UNITE, the antiSMASH databases, the Tiberius model, the telomere repeat database, the Illumina adapter records of NCBI UniVec, the PhiX genome, the BUSCO lineage data, the Medaka model weights and, with the resistance-mutation screen on, ChroQueTas with its FungAMR database — is downloaded on the first run that needs it, most of it once per output folder, or all at once with snakemake download_all. directories.unite_db and directories.antismash_db take copies you already hold. Download commands, disk sizes and every optional key are on Reference databases. Running on a cluster whose compute nodes have no internet is described on On a cluster.

Output

Everything lands under directories.output_dir, in numbered folders that are the same in every mode: 00.resources for material shared by the samples of a run, then 01.reads (not in contigs mode) through 06.report, and logs. The delivered genome is always 02.assembly/{sample}/contigs_final.fasta. The annotation — GenBank, GFF3, proteins and the functional annotation as a table with one row per transcript — is in 04.annotation/{sample}/funannotate/annotate_results/, and the QC of all samples in 06.report/multiqc_report.html.

funannotate renames the contigs scaffold_1, scaffold_2 and so on, while QUAST, BUSCO and the ITS files keep the original names. The funannotate database is unpacked into directories.funannotate_db, outside the output folder, and Snakemake keeps its records and the conda environments in .snakemake/ in the directory you launch from. One fungal genome in contigs mode takes several hours and leaves in the order of 10 GB. What each stage holds, file by file, is on Output files.

Citation

Antonielli, L., & Brader, G. (2026). FunFlux: Integrated workflow for fungal genome assembly and annotation (v2.1.0). Zenodo. https://doi.org/10.5281/zenodo.13612159

That DOI is the concept DOI and always resolves to the newest release; each tagged release also gets its own version DOI, which is the one to cite when the exact code matters.

Most of the results FunFlux reports are produced by other people's tools and databases, so cite those too. The list is on Citation and references.

Acknowledgements

This work was supported by the RATION project (Risk AssessmenT InnOvatioN for low-risk pesticides), funded by the European Union's Horizon Europe research and innovation programme under grant agreement No. 101084163.

Much of v2.0.0 — merging FunFlux 1.x, Funnotator and FunFluxL into one workflow, adding the hybrid mode and writing the documentation — was done with Claude Code. Anthropic provided six months of Claude Max through their Open Source programme, and the scale of the v2 rewrite would not have been realistic without it.

Licence

FunFlux is released under the MIT License. The repository ships code, download addresses and two small files of adapter sequences (from Oxford Nanopore and NCBI UniVec), and no databases: every database is either one you already hold or one the workflow fetches from its publisher, under that publisher's terms. The tools and databases FunFlux runs carry their own licences, listed with their conditions on Licensing. GeneMark-ES, SignalP 6, Medaka, ChroQueTas with its FungAMR database (the resistance-mutation screen, off by default), and the PROSITE and SMART databases that InterProScan searches in every run restrict commercial use.