Coming from v1

This page explains how to move from FunFlux 1.x, Funnotator, FunFluxL or FunFluxL+ to FunFlux v2. Version 2.0.0 replaces these four workflows, and the mode key selects what used to be a separate workflow.

Terms used on this page

NOTE	a line starting [FunFlux] NOTE that reports something to check without stopping the run; read them after every dry run
launch folder	the folder Snakemake is started from, normally the FunFlux repository; relative paths are read from it, and .snakemake/ is created there
funannotate database	the frozen snapshot of funannotate's data, such as Swiss-Prot, Pfam, the BUSCO set and stored AUGUSTUS models, downloaded once from Zenodo into directories.funannotate_db
path repair	rule normalize_funannotate_db, which rewrites the absolute paths stored in the funannotate database to your folder, in place and without a lock
the screen	the contamination and mitochondrion screen of the draft: BLAST against core_nt, BlobTools, the mitochondrion screen and the contig selector; in contigs mode only with parameters.contigs.screen: true
confirmed mitochondrial contig	a contig the mitochondrion screen classes confirmed_mitochondrial; with the default parameters.mitochondria.mode: separate_confirmed, it is kept whatever its genus
guide assembly	in hybrid mode, the SPAdes assembly of the Illumina reads; it is screened, and only the pairs that map to it go on

Was	Input	Now
FunFlux 1.x	Illumina paired-end reads	mode: illumina
Funnotator, shipped inside FunFlux 1.x as workflow/Funnotator	assembled genomes (FASTA)	mode: contigs
FunFluxL 0.1	ONT or PacBio HiFi reads	mode: longread
FunFluxL+, never published	Illumina pairs plus ONT or PacBio HiFi reads	mode: hybrid

The FunFlux repository continues as v2 under the same concept DOI, 10.5281/zenodo.13612159. Cite results produced with a v1 workflow by the version you ran, as described on Citation and references.

The v1 workflows are no longer in the main branch, but their releases stay available. FunFlux v1.1.0, with Funnotator inside it, is the tag v1.1.0 of this repository: git clone --branch v1.1.0 https://github.com/iLivius/FunFlux.git FunFlux-v1.1.0 puts it in a folder of its own. FunFluxL 0.1.0 is the tag v0.1.0 of the FunFluxL repository.

Neither a v1 configuration nor a v1 output folder can be reused as it is. Make these changes, in this order:

1. Start from the v2 template, add mode and the input: block, and move the long-read settings into parameters.longread or parameters.hybrid (Why a v1 configuration stops).
2. Name the configuration with --configfile, give --cores the value of resources.threads, and, on a single computer, drop --jobs (Command line).
3. Give v2 a new, empty output_dir (Output folders).
4. Point directories.funannotate_db at a new, empty folder, to get the newer snapshot of the funannotate database (Behaviour changes).
5. Replace decontamination.mode: auto and parameters.genus, if you used them (Behaviour changes).
6. Update scripts that read renamed files or columns (Output folders, Behaviour changes).

The "What to do" column of Behaviour changes lists the rest.

Why a v1 configuration stops

A v1 configuration has no mode key, so FunFlux v2 stops before any job starts:

[FunFlux] mode: this configuration has no 'mode' key. FunFlux v2 needs one of illumina | longread | hybrid | contigs, and an 'input:' block that names the input directory for that mode. A configuration written for FunFlux, FunFluxL or FunFluxL+ v1 does not run unchanged: start from the config template shipped with FunFlux v2 (see 'Coming from v1' in the documentation).

Adding mode: is not enough. The input folder moved out of directories, and FunFluxL and FunFluxL+ kept their long-read settings at the top of parameters, where v2 stops on them:

[FunFlux] mode=illumina requires input.illumina_dir to be set in the config.
[FunFlux] parameters.flye_input_mode: this setting moved into the block named after your mode: parameters.longread.flye_input_mode or parameters.hybrid.flye_input_mode.

Start from the v2 template and copy your paths and settings into it:

cp config/config.yaml config/config_custom.yaml

The table lists the v1 keys that moved or were renamed, and some new keys. The other new keys are named under Behaviour changes.

v1 key	v2 key	Note
none	mode	required: illumina, longread, hybrid or contigs
directories.input_dir	input.illumina_dir, input.longread_dir or input.contigs_dir	hybrid reads two keys, which may name the same folder; v2 ignores directories.input_dir without a message
none in FunFluxL 0.1, which assembled with Flye; parameters.long_read_assembler in FunFluxL+	parameters.<mode>.assembler	hifiasm (the default) or flye
parameters.flye_input_mode	parameters.<mode>.flye_input_mode	the value pacbio-hifi is gone: name the reads {sample}_pacbio.<ext>
parameters.medaka_model	parameters.<mode>.medaka_model
none in FunFluxL 0.1, which ran Filtlong with --min_length 1000 --keep_percent 90 --target_bases 500000000; parameters.filtlong.* in FunFluxL+	parameters.<mode>.filtlong.*	min_length 1000, keep_percent 90, length_weight 10, target_bases empty (no cap); see Behaviour changes
parameters.filtlong.split (FunFluxL+)	parameters.hybrid.filtlong_split	a split key inside filtlong stops the run
parameters.hifiasm.* (FunFluxL+)	parameters.<mode>.hifiasm.*
parameters.genus	parameters.decontamination	translated with a NOTE when there is no decontamination block: a genus becomes mode: include, an empty value mode: off (v1: mode: auto, which v2 does not have)
none (FunFlux 1.x searched core_nt)	parameters.nt_version	now read in every mode
none (FunFlux 1.x kept 10 database sequences per contig, with 1 alignment each)	parameters.blast_max_target_seqs, parameters.blast_max_hsps	defaults 10 and 5

On this page, <mode> is longread or hybrid. The long-read settings live in the block named after the mode, and only that block is read. All other v1 keys keep their names and meanings, apart from the exceptions listed under Behaviour changes, such as decontamination.mode: auto, which is gone. In addition, 'null' in quotes as medaka_model means off, as described on The Medaka model.

Command line

The table compares the v1 and v2 command lines. Relative paths and the conda environments now belong to the launch folder, not to output_dir.

	v1	v2
Command	snakemake --configfile config/config.yaml --sdm conda --cores 12 --jobs 2; Funnotator added --snakefile workflow/Funnotator	snakemake --sdm conda --cores 24 --configfile config/config_custom.yaml
Configuration	config/config.yaml was loaded by default	must be named with --configfile; without it FunFlux stops with [FunFlux] No configuration supplied.
CPUs	most rules took their threads from resources.threads through resources: cpus, which --cores does not limit, so tools could run on more threads than --cores allowed	every CPU-bound rule declares threads:, so --cores, equal to resources.threads, limits the whole run; no --jobs
Working folder	Snakemake moved into output_dir (workdir:): relative paths in the config were read from there, and the conda environments were built in <output_dir>/.snakemake/conda/	Snakemake stays in the launch folder: relative paths are read from there, and .snakemake/ with the conda environments is created there

Output folders

FunFlux v2 does not upgrade a v1 output folder in place

Give v2 a new output_dir. Do not resume a v1 run with v2, and do not point v2 at a v1 output folder. The old layout is not converted, and some paths exist in both layouts (02.assembly/{sample}/, 04.annotation/{sample}/), so old and new files would mix.

The v2 layout is the same in every mode, and a mode that has nothing for a folder leaves it out:

<output_dir>/
├── 00.resources/    material shared by the run: UNITE, antiSMASH databases, telomere repeat database, SignalP 6, ChroQueTas, PhiX
├── 01.reads/        read QC, filtered and selected reads (not in contigs mode)
├── 02.assembly/     assembly, screen, mitochondrion split, QUAST, BUSCO, Qualimap
├── 03.taxonomy/     ITS extraction and classification
├── 04.annotation/   masking, gene prediction, functional annotation
├── 05.amr/          antifungal resistance mutations (new; only with the AMR screen on)
├── 06.report/       MultiQC
└── logs/            one log per rule and sample

FunFlux 1.x and FunFluxL

v1	v2
01.pre-processing/ (PhiX index, fastp reports, trimmed reads)	00.resources/phix/, 01.reads/{sample}/illumina/
01.pre-processing/{sample}_long_filt.fastq, 01.pre-processing/read_qc/ (FunFluxL)	01.reads/{sample}/long/
02.assembly/{sample}/	02.assembly/{sample}/
03.post-processing/contaminants/{sample}/, and the read-mapping BAM in 03.post-processing/{sample}/ (FunFlux 1.x) or 03.post-processing/contaminants/{sample}_map.bam (FunFluxL)	02.assembly/{sample}/contaminants/
03.post-processing/consensus/{sample}/ (FunFluxL, Medaka)	02.assembly/{sample}/medaka/
03.post-processing/assembly_evaluation/{sample}/ (FunFlux 1.x)	02.assembly/{sample}/eval/quast/, now also in longread mode
03.post-processing/completeness_evaluation/{sample}/	02.assembly/{sample}/eval/busco/
03.post-processing/mapping_evaluation/{sample}/	02.assembly/{sample}/eval/qualimap/
03.post-processing/ITS_extraction/{sample}/	03.taxonomy/{sample}/
03.post-processing/ITS_extraction/unite_its_sintax.fasta	00.resources/unite/
04.annotation/{sample}/	04.annotation/{sample}/
04.annotation/antismash_db/	00.resources/antismash/databases/
report/	06.report/

FunFluxL+

FunFluxL+	v2
01.read_processing/illumina/, 01.read_processing/long/	00.resources/phix/, 01.reads/{sample}/illumina/, 01.reads/{sample}/long/
02.short-read_assembly/{sample}/	02.assembly/{sample}/guide/
03.read_selection/contamination/{sample}/	02.assembly/{sample}/guide/contaminants/; the selected pairs in 01.reads/{sample}/selected/
03.read_selection/mapping_evaluation/{sample}/	02.assembly/{sample}/guide/eval/qualimap/
03.read_selection/assembly_evaluation/{sample}/ and 07.genome_qc/assembly_evaluation/{sample}/	02.assembly/{sample}/eval/quast/, one report with the labels {sample}_guide and {sample}
04.long-read_assembly/{sample}/	02.assembly/{sample}/ (hifiasm/ or flye/, assembly.fasta)
05.assembly_curation/contamination/{sample}/	02.assembly/{sample}/contaminants/
05.assembly_curation/mapping_evaluation/{sample}/	02.assembly/{sample}/eval/qualimap/
06.polishing/medaka/{sample}/, 06.polishing/polypolish/{sample}/	02.assembly/{sample}/medaka/, 02.assembly/{sample}/polypolish/
07.genome_qc/completeness_evaluation/{sample}/	02.assembly/{sample}/eval/busco/
07.genome_qc/ITS_extraction/{sample}/	03.taxonomy/{sample}/
07.genome_qc/ITS_extraction/unite_its_sintax.fasta	00.resources/unite/
08.annotation/{sample}/, 08.annotation/antismash_db/	04.annotation/{sample}/, 00.resources/antismash/databases/
09.report/	06.report/

Funnotator

Funnotator	v2
01.pre-processing/pre_filtering/{sample}/contigs_filt.fasta	02.assembly/{sample}/contigs_final.fasta
01.pre-processing/assembly_evaluation/{sample}/	02.assembly/{sample}/eval/quast/
01.pre-processing/completeness_evaluation/{sample}/	02.assembly/{sample}/eval/busco/
01.pre-processing/ITS_extraction/{sample}/	03.taxonomy/{sample}/
01.pre-processing/ITS_extraction/unite_its_sintax.fasta	00.resources/unite/
02.annotation/{sample}/, 02.annotation/antismash_db/	04.annotation/{sample}/, 00.resources/antismash/databases/
report/	06.report/

The delivered assembly files have new names, and the mitochondrion now has a file of its own:

v1	v2, in 02.assembly/{sample}/
contigs_sel.fasta (FunFlux 1.x), the Medaka consensus.fasta (FunFluxL), polished.fasta (FunFluxL+)	contigs_all.fasta: the complete assembly after the screen and polishing, mitochondrion included
none	contigs_final.fasta: the nuclear genome, read by QUAST, BUSCO, the ITS steps and funannotate
none	contigs_mitochondrial.fasta: the confirmed mitochondrial contigs; an empty file when there are none

A new file, 04.annotation/{sample}/contigs_removed_by_clean.txt, lists the contigs that funannotate clean removed before gene prediction.

Behaviour changes

The tables below group the changes by part of the workflow. The "Affects" column names the v1 workflows that a change concerns, and the "What to do" column names the action. "Nothing" means that the change needs no action.

Running and configuration

Change	Affects	What to do
One workflow with a mode key; the configuration must be named with --configfile	all	Start from the v2 template; add mode and the input block
Every CPU-bound rule declares threads:, so --cores limits the whole run	all	Give --cores the value of resources.threads (24 in the template, as in v1); drop --jobs on a single computer
FunFlux runs on a cluster with --executor slurm --jobs N. Every heavy step asks the scheduler for resources.ram_gb of memory and for the time limit of the new key resources.runtime_hours (default 48). Every download step runs where Snakemake runs, and snakemake download_all fetches all downloads in advance (On a cluster)	all	On a cluster, set the three resources keys from the limits of a node and a partition
curl, tar and gzip come from a conda environment (download), and InterProScan's Perl from the interproscan environment; v1 took them from the computer	all	Nothing
No workdir:: relative paths are read from the launch folder	all	Use absolute paths, or paths relative to the launch folder
Thread caps: eggNOG-mapper 12 (FunFlux 1.x: 24), nhmmer and ITSx 8, SINTAX 1, SignalP 6 at most resources.threads (CPU cores and threads)	all	Nothing
Conda environments built from pinned package lists of the tested tool versions (workflow/envs/*.linux-64.pin.txt), with the Python version pinned in every environment that contains Python	all	Nothing; the first v2 run builds new environments
New optional keys whose defaults keep the v1 behaviour: spades_kmers, busco.lineage, prediction.busco_db, repeatmodeler_seed, directories.antismash_db, directories.unite_db, contigs.screen, medaka_model_fallback_auto	all	Nothing

Inputs

Change	Affects	What to do
One input key per technology: input.illumina_dir, input.longread_dir, input.contigs_dir replace the v1 key directories.input_dir	all	Move the path
The suffixes stay _ont and _pacbio, and flye_input_mode: pacbio-hifi is removed. Illumina plus PacBio HiFi is allowed in hybrid	FunFluxL, FunFluxL+	Name PacBio HiFi reads {sample}_pacbio.<ext>
Underscores are allowed in sample names (v1 rejected them). FunFlux prints a NOTE when a name is not a valid locus-tag prefix	all	For GenBank submission, use 3 to 12 letters or digits starting with a letter

Assembly and screen

Change	Affects	What to do
Long reads are assembled with hifiasm by default. FunFluxL 0.1 used Flye, which stays available	FunFluxL, FunFluxL+	Set parameters.<mode>.assembler: flye to keep Flye
Since FunFlux 2.1.0, the ONT adapter and the barcode after it are trimmed from the reads of ONT samples with fastplong before Filtlong. By default (parameters.<mode>.ont_adapter: detect) FunFlux works out from the reads what to trim, and leaves reads that are already trimmed as they are. FunFluxL 0.1 and FunFluxL+ assembled the reads untrimmed (Adapter trimming)	FunFluxL, FunFluxL+, ONT samples	Nothing; ont_adapter: off assembles the reads untrimmed, as v1 did
Filtlong keeps every read that passes its length and quality filters: target_bases is empty (no cap) by default, and length_weight: 10 favours longer reads. FunFluxL 0.1 kept at most 500 Mb (--target_bases 500000000) with Filtlong's default length weight of 1	FunFluxL	To repeat FunFluxL 0.1's read selection, set parameters.longread.filtlong.target_bases: 500000000 and length_weight: 1
Before the assembly starts, a named Medaka model is checked against Medaka's list of models, and auto is worked out from the basecaller tag of the first 100 reads of the raw FASTQ. A wrong model therefore fails in seconds instead of after the assembly. auto needs network access once and exactly one basecaller tag in those reads. New key medaka_model_fallback_auto. The assembler waits for the check's file, so a changed medaka_model also re-runs the assembly of the ONT samples	FunFluxL, FunFluxL+, ONT samples	For public or re-headered ONT reads, name the model or set medaka_model: false (The Medaka model)
The PhiX genome and its bowtie2 index are downloaded once per output folder and kept in 00.resources/phix/. In v1, they were kept with the processed reads and deleted after the run. The PhiX-free pairs are called {sample}_nophix_R1.fastq and _R2.fastq	FunFlux 1.x, FunFluxL+	Nothing
SPAdes writes into 02.assembly/{sample}/spades/ (in hybrid, guide/spades/), so a re-run of SPAdes no longer deletes the screen's files of the sample. In FunFlux 1.x the whole sample folder was SPAdes' output. Flye likewise works in flye/, and its assembly is copied up one level	FunFlux 1.x, FunFluxL, FunFluxL+	Nothing
Filtlong guidance by the Illumina reads is off by default. Switched on (short_read_guidance: true, filtlong_split: 1000), it applies to ONT samples only. FunFluxL+ always guided and cut every sample, PacBio included	FunFluxL+	For FunFluxL+'s behaviour on ONT samples, set short_read_guidance: true; move filtlong.split to parameters.hybrid.filtlong_split (Filtlong guidance)
Polypolish output headers are cut at the first space or tab (v1: >ptg000001l polypolish)	FunFluxL+	Nothing
assembly_linearized.fasta and assembly_single_line.fasta are no longer written	FunFluxL, FunFluxL+	Update scripts that read them
Since FunFlux 2.1.0 the BLAST search keeps 10 database sequences per contig, with up to 5 alignments each, in every mode, and drops matches with an e-value above 1e-10 (FunFlux 2.0.0: 50, 5 and 1e-5; FunFlux 1.x: 10, 1 and 1e-5; FunFluxL and FunFluxL+: 50, 5 and 1e-5). FunFlux prints a NOTE when blast_max_hsps is below 5 and the mitochondrion screen is on	FunFlux 1.x, FunFluxL, FunFluxL+	Nothing. To come close to earlier numbers, set blast_max_hsps: 1 (FunFlux 1.x) or blast_max_target_seqs: 50 (FunFluxL, FunFluxL+) (BLAST hit depth)
parameters.nt_version is read in illumina mode too	FunFlux 1.x	Set it if your BLAST database has another name than core_nt
New: a mitochondrion screen and split in every mode with the screen. The screen classes each contig of the draft from its BLAST hits and, in longread and hybrid, from the assembler's read depth and circularity. With the default parameters.mitochondria.mode: separate_confirmed, the confirmed mitochondrial contigs are kept whatever their genus, on the hybrid guide assembly too. After polishing they are split off into contigs_mitochondrial.fasta, so that QUAST, BUSCO, the ITS steps and funannotate read only the nuclear genome, contigs_final.fasta. Since FunFlux 2.1.0, a contig whose best mitochondrial match is an animal's or a plant's, such as human mitochondrial DNA from the library, is not confirmed (foreign_mitochondrial): it stays in the assembly, where the genus filter treats it like any other contig. A confirmed mitochondrial contig missing from the BlobTools table is kept as no-taxonomy, and a confirmed ID missing from the FASTA stops the run (Mitochondria)	all	Read contigs_final.fasta as the genome; parameters.mitochondria.mode: report or off keeps the mitochondrial contigs in it
decontamination.mode: auto is removed and stops the run before any job starts, also as a mode in sample_overrides when the screen runs. BLAST genus labels are unreliable for poorly sequenced fungi, so auto can remove most of the genome (Configuration)	FunFlux 1.x, FunFluxL, FunFluxL+	Use off (the default), read contig_taxonomy_decisions.tsv, then exclude the contaminant genera or include the genera BLAST assigned to the isolate's own contigs
With the screen, the decontamination files (include_genera_by_sample, exclude_genera_file, sample_overrides) and an include mode without a genus are checked before any job starts (v1: when select_contigs ran, after BLAST)	FunFlux 1.x, FunFluxL, FunFluxL+	Nothing
parameters.genus is translated with a NOTE, and an empty genus now means mode: off (v1: auto, the genus with the most contigs). The template sets discard_no_hit: false, where the v1 template had true (the v1 code default was already false)	all	Replace genus with parameters.decontamination; set discard_no_hit: true if you relied on the v1 template
With SPAdes headers, the contigs mode filter keeps contigs of at least 500 bp and 2x coverage. Any other FASTA only has its headers cut at the first space or tab (Funnotator removed every contig of 500 bp or less). An empty result stops the run, and the screen is optional	Funnotator	Expect 500 bp contigs to be kept and no length filter on non-SPAdes FASTA
Delivered files renamed (table above)	all	Update downstream scripts

QC and ITS

Change	Affects	What to do
QUAST also runs in longread mode. In hybrid, one QUAST report holds the guide ({sample}_guide) and the final assembly ({sample})	FunFluxL, FunFluxL+	Nothing
BUSCO's result tree is now kept, except hmmer_output/ and busco_sequences/ (v1 deleted run_*/ and auto_lineage/). With the auto lineage, that is up to about 2 GB per sample. With auto, the lineage data is still downloaded per sample into a temporary folder, while a named lineage is downloaded once per output folder into 00.resources/busco/ and read offline. New key busco.lineage. With a named lineage, MultiQC shows one BUSCO row per sample instead of two	all	Allow up to about 2 GB per sample, or name a lineage, for example hypocreales_odb12, which keeps the result tree of that lineage only
New: the contig summary, 02.assembly/{sample}/eval/contig_summary.tsv. It has one row per contig of the draft assembly, with a verdict (chromosome, chromosome_arm, mitochondrion, rDNA_array, low_depth, repeat, nuclear_contig, nuclear_fragment, contaminant, foreign_mitochondrion or unclassified) and the evidence behind it, and is also shown as a table in the MultiQC report. Its annotated_as column links each contig to its scaffold_N name in the annotation (Contig summary). The telomere repeat unit of each assembly is chosen from the data with tidk and checked against tidk's telomere repeat database, downloaded once per output folder. New keys parameters.contig_summary.*, new required link links.telomere_db_link	all	Copy the links.telomere_db_link line from the v2 template; the run stops without it
New since FunFlux 2.1.0: the contig adapter screen. Every draft assembly is searched with BLAST for ONT, PacBio, MGI and Illumina adapters and barcodes; the loci go to 02.assembly/{sample}/eval/adapter_screen.tsv, to the adapter column of the contig summary and to the Adapters table of the MultiQC report. The Illumina records of NCBI UniVec are downloaded once per output folder. New optional link links.univec_link and key parameters.contig_summary.univec_sha256 (The contig adapter screen)	all	Nothing; the screen only reports, and changes no genome file
ITS is cut out of contigs_final.fasta in one strand-oriented window per rDNA unit before ITSx runs (Funnotator ran ITSx on its raw input). New keys its.window_flank (2,000 bp on each side of the 5.8S gene) and its.max_full_length (1,500 bp). Identical ITS copies are collapsed and counted. ITS1 or ITS2 outside 50-600 bp is listed in {sample}_ITS_copies.txt and not classified, and a full ITS longer than its.max_full_length is marked rejected_too_long there (only ITS1 and ITS2 are classified, as in v1). Long-read assemblies, which could give empty or wrong ITS files in v1, now give the ITS	all	Update scripts that parse ITS record names: records are named after their window (contig:start-end_strand), classified copies {sample}_ITS1_1, {sample}_ITS2_1, ... (ITS on long-read contigs)
SINTAX runs on one thread with --randseed 1, so the ITS confidences repeat from run to run. In v1, SINTAX used several threads and no fixed seed	all	Nothing
SINTAX classifies against the UNITE records whose taxonomy names a genus (new rule unite_identified, new key its.unite_identified_only, default true). In v1, SINTAX used the whole release, in which about a third of the records stop above the genus, and one identical to the query can stop the result at d:Fungi	all	Nothing; its.unite_identified_only: false gives the v1 reference
QC-only rules (QUAST, Qualimap, NanoPlot) no longer stop the run when the tool fails	all	Check the log and the MultiQC page for a missing table
NanoPlot writes its plots as HTML pages only (--no_static). FunFluxL also wrote them as PNG, but the pinned NanoPlot 1.46.2 draws PNG only where a Chrome or Chromium browser is installed, so the output would differ between computers	FunFluxL, FunFluxL+	Open the HTML plots; MultiQC reads the NanoStats table as before
Qualimap runs without a display, so it works on a server with no X session. Its Java heap is min(ram_gb, 64) GB, as in v1	FunFlux 1.x, FunFluxL, FunFluxL+	Nothing
BUSCO runs with --opt-out-run-stats, so it sends no usage statistics and makes no ipinfo.io lookup (v1 sent both). MultiQC no longer shortens sample names, so a sample named SAMPLE1.clean keeps its name	all	Nothing

Annotation and databases

Change	Affects	What to do
The antiSMASH database stays in 00.resources/antismash/databases/ between runs on the same output folder (v1 deleted it and downloaded it again on every resume), or is linked from directories.antismash_db. The same holds for UNITE, with PROVENANCE.txt	all	Optionally point directories.antismash_db and directories.unite_db at copies you have
The funannotate database archive is checked with sha256sum -c before it is unpacked (v1 downloaded the checksum file but did not check it). A mismatch stops the run	all	Delete the archive and its .sha256 file and run again
The template links a newer snapshot of the funannotate database, funannotate_db_1.8.17_2026-10-08 (Zenodo record 23265189), with newer Swiss-Prot, Pfam, InterPro, Gene Ontology and gene2product releases. The v1 templates link the 2025-12-15 snapshot (record 18271295). What changed is listed under funannotate database snapshot	all	Point directories.funannotate_db at a new, empty folder, because a folder that holds the v1 snapshot keeps being used
Since FunFlux 2.1.0, InterProScan runs with -dp: it computes every match itself and does not ask EBI's lookup service for matches already computed (the service did not answer InterProScan 5 in November 2026), so it needs no internet connection (Functional annotation)	all	Nothing
With GeneMark-ES on, the new rule genemark_check runs GeneMark-ES on a 200 kb slice of each masked genome before gene prediction (about a minute). An expired licence key therefore stops the run before funannotate predict starts instead of partway through it	all, with GeneMark-ES	Renew the key when the message asks for it
The AUGUSTUS model (column 4 of the annotation parameter table, files.annotation_params) is optional. Empty or - makes funannotate train AUGUSTUS on the genome's BUSCO genes, which v1 never did, because v1 required a model. A named model must exist in the funannotate database, or the run stops with the closest names (v1 passed any name on, and funannotate silently trained AUGUSTUS under a misspelt one). The new rule augustus_model_check records the choice in 04.annotation/{sample}/augustus_model.txt and at the top of the gene prediction log. parameters.prediction.augustus_optimize adds the long training	all	Nothing for a v1 table with a valid model name: AUGUSTUS gets the same option as in v1
Tiberius, a gene finder with a model trained by its authors on many fungal genomes, runs by default before gene prediction, and funannotate predict adds its gene models to the evidence it combines (--other_gff, weight 1). It needs about 16 GB of memory and downloads its model (30 MB) once per output folder. New keys parameters.tiberius.mode and model_sha256, new link links.tiberius_model_link	all	Nothing; parameters.tiberius.mode: off gives gene prediction without it, as in v1
New key parameters.genemark.mode: auto, the default, uses GeneMark-ES when directories.genemark_dir is set, and off never uses it. An empty value (mode: or mode: "") means auto, as for the other mode keys	all	Nothing; leave directories.genemark_dir empty to run without GeneMark-ES
New, off by default: SignalP 6, which predicts signal peptides for funannotate annotate. It needs the licensed archive from DTU (files.signalp6_archive) and a parameters.signalp6.mode other than off (Licensed and external tools)	all	Nothing
New: rule normalize_funannotate_db, the path repair, rewrites the absolute paths stored inside the funannotate database to your folder. It runs in place and without a lock, so let one run pass it before others start on the same database (Reference databases)	all	Nothing
New, off by default: the AMR screen. ChroQueTas screens each genome for the point mutations that the FungAMR database lists with evidence for or against resistance, and writes 05.amr/{sample}/. New keys parameters.amr.*, new link links.chroquetas_link (Antifungal resistance mutations)	all	Nothing; parameters.amr.chroquetas: on switches it on, and accepts the non-commercial licence of ChroQueTas

Logs

Change	Affects	What to do
One log per rule and sample in logs/; InterProScan's scratch folder is 04.annotation/{sample}/iprscan/tmp/ (no top-level temp/)	all	Nothing

Kept as in v1

• decontamination.mode: off as the default
• BUSCO on the final genome only
• the filtered long reads, the selected pairs and the read-mapping BAM of the read-based modes, kept after the run