Assembly QC¶

Three tools check the quality of each sample's assembly. QUAST measures its size and contiguity, and BUSCO its completeness. In the read-based modes, Qualimap also shows how the sample's reads map back to the draft. Read this page to interpret their results, to choose the BUSCO lineage of a batch and to find their files.

QUAST and BUSCO run on 02.assembly/{sample}/contigs_final.fasta, the nuclear genome that ITS extraction and funannotate also read. When the contamination and mitochondrion screen runs with the default parameters.mitochondria.mode: separate_confirmed, the mitochondrion has been split off before this point, as described in Mitochondria. If it were left in, its genetic code, gene content and copy number would distort both the contig statistics and the BUSCO counts of a genome that is annotated as nuclear. In contigs mode without the screen, contigs_final.fasta is the input FASTA after the contig filter.

Tool	Version	Measures	Rule	Output, in `02.assembly/{sample}/`	If the tool fails
QUAST	5.3.0	size and contiguity	`genome_assembly_evaluation`	`eval/quast/`	a line in the log; the run continues
BUSCO	6.0.0	completeness of conserved single-copy genes	`completeness`	`eval/busco/`	the job fails and the run stops (`--keep-going` lets the other jobs finish)
Qualimap	2.3	reads mapped back to the draft (read-based modes)	`map_evaluation`	`eval/qualimap/`	a line in the log; the run continues

Only MultiQC and the contig summary read these results. The contig summary counts the complete BUSCO markers on each contig and is written after the annotation (Contig summary).

Terms used on this page


nuclear genome	the delivered genome without the confirmed mitochondrial contigs, `contigs_final.fasta`; it holds every contig when nothing is split off (`parameters.mitochondria.mode` `off` or `report`, or no screen)
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`
draft assembly (the draft)	the assembly the screen examines, before any contig is removed: `contigs_filt.fasta` in `illumina` mode and in `contigs` mode with the screen, `assembly.fasta` in `longread` and `hybrid`
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
read-based modes (read modes)	the modes `illumina`, `longread` and `hybrid`, which start from reads, as against `contigs`
coverage track	the reads aligned back to the draft, `contaminants/{sample}_map.bam`; BlobTools takes each contig's coverage from it, the contig summary its read depth and Qualimap its report; in `contigs` mode, the contigs aligned to themselves
lineage dataset	the BUSCO gene set of one taxonomic group, such as `hypocreales_odb12`, for the completeness check; with `lineage: auto`, BUSCO chooses one by placing the genome
contig summary	the table `02.assembly/{sample}/eval/contig_summary.tsv`: one row per contig of the draft, with its evidence (telomeres, depth, rDNA units, genes) and a verdict

QUAST¶

QUAST measures the size and contiguity of the genome. Rule genome_assembly_evaluation runs it:

quast contigs_final.fasta -o eval/quast --labels {sample} --no-icarus -t <threads>
# hybrid: the screened guide assembly first, then the final genome, two labels
quast guide/contaminants/contigs_decontam.fasta contigs_final.fasta -o eval/quast \
  --labels {sample}_guide,{sample} --no-icarus -t <threads>

QUAST uses min(resources.threads, 24) threads. It runs without a reference genome, so it reports size and contiguity only, with no misassemblies, no genome fraction and no NG50. Contigs shorter than 500 bp, QUAST's default minimum, are left out of every statistic except the rows marked (>= 0 bp). --labels names the assembly after the sample, so QUAST and MultiQC show the sample name. --no-icarus leaves out QUAST's interactive contig viewer.

When QUAST fails, its log ends with [FunFlux] genome_assembly_evaluation: QUAST exited with an error for sample S. ..., eval/quast/ stays in place, and the MultiQC page has no QUAST row for that sample.

In hybrid mode, one QUAST call compares the screened Illumina guide assembly with the final genome, labelled {sample}_guide and {sample} in that order. The guide is the short-read view of the same isolate, so the two rows show what the long reads added.

What the numbers mean¶

QUAST writes its report as eval/quast/report.tsv, also as .txt, .html, .pdf and .tex, and transposed as transposed_report.*. These are some of its rows:

Row	Meaning
`# contigs (>= 0 bp)`	all contigs
`# contigs (>= 1000 bp)` to `(>= 50000 bp)`	contigs of at least 1, 5, 10, 25 and 50 kb
`Total length (>= 0 bp)`	sum of all contig lengths (bp)
`Total length (>= 50000 bp)`	bases in contigs of 50 kb or more
`# contigs`	contigs of 500 bp or more
`Largest contig`	length of the longest contig (bp)
`Total length`	bases in contigs of 500 bp or more
`GC (%)`	GC content
`N50`	contigs of this length or longer hold half of the total length
`N90`	the same for 90 % of the total length
`auN`	sum of the squared contig lengths divided by the total length: the mean length of the contig that a random base lies in
`L50`	the number of contigs, longest first, that hold half of the total length
`L90`	the same for 90 %
`# N's per 100 kbp`	gap characters

A higher N50 and auN and a lower L50 mean fewer and longer contigs.

BUSCO¶

BUSCO searches the genome for genes that nearly every member of a lineage carries exactly once (single-copy orthologs) and reports how many it finds whole, duplicated, in pieces or not at all.

busco --mode genome -i contigs_final.fasta --out {sample} --out_path eval/busco \
  --download_path eval/busco/busco_tmp --auto-lineage-euk --opt-out-run-stats --cpu <threads>

Rule completeness uses min(resources.threads, 24) threads, as QUAST does. With a fixed lineage, -l <lineage> replaces --auto-lineage-euk, and --offline --download_path 00.resources/busco replaces the download folder of the sample. In a eukaryote genome, BUSCO 6.0.0 finds the genes with miniprot unless another predictor is requested. FunFlux requests none, and the summary says BUSCO was run in mode: euk_genome_min.

Automatic or fixed lineage¶

The key parameters.busco.lineage names the dataset BUSCO uses, or leaves the choice to BUSCO:

parameters:
  busco:
    lineage: auto        # or a dataset name such as hypocreales_odb12

Setting	What BUSCO runs	Summaries in `eval/busco/`	MultiQC rows per sample
`auto` (default)	eukaryota_odb12, then the most specific lineage it reaches by placing the genome in the eukaryote tree	`short_summary.generic.eukaryota_odb12.{sample}.txt` and `short_summary.specific.<lineage>.{sample}.txt`	2
a dataset name	that dataset only	`short_summary.specific.<lineage>.{sample}.txt`	1

For a batch, a fixed lineage keeps the BUSCO numbers comparable

Each sample is placed on its own, so auto can choose different lineages for the samples of one batch, and their numbers are then not comparable. A fixed lineage gives every sample the same dataset, skips the eukaryota run and the placement, and leaves a smaller result folder.

Each summary also exists as .json. The log names the lineage that the placement chose, in the line Lineage <lineage> is selected, supported by <n> markers out of <m>. A dataset name must match [a-z0-9_]+_odb[0-9]+ after lower-casing, or the dry run stops. busco --list-datasets prints the available names, each with its version after a dot, as in hypocreales_odb12.2. Leave out the dot and the number after it.

Reading the summary¶

Each summary has one result line of this form:

C:<C>%[S:<S>%,D:<D>%],F:<F>%,M:<M>%,n:<n>,E:<E>%

C is the completeness of the genome, and D, F, M and E say what lies behind it:

Letter	Meaning	What an unusual value can mean
C	complete: found over its full length, S + D	a lower C after a decontamination policy removed contigs points to removed genome rather than removed contamination (Decontamination)
S	complete, one copy
D	complete, two or more copies	a high D can come from a recent whole-genome duplication, from haplotypes assembled separately, or from a mixed culture of two fungi of the same lineage, in which each gene is present twice
F	fragmented: found in part	the gene is only partly present, or no full-length gene model was built although the gene may be there
M	missing	the gene is absent, or the search found no significant match
n	genes in the dataset
E	complete genes whose miniprot prediction contains an internal stop codon	caution before using BUSCO's predicted sequences downstream; those genes still count as complete

The readings of D (genome duplication, haplotypes), F, M and E are those of the BUSCO user guide.

The header of each summary names the dataset and its creation date, for example hypocreales_odb12 (Creation date: 2026-05-22, number of genomes: 76, number of BUSCOs: 4349). With auto, BUSCO downloads the current dataset at every run, and a newer dataset can change the numbers. Compare samples only across runs with the same dataset date. A fixed lineage is downloaded once per output folder, so every later run into that folder uses the same release.

Downloads, disk and network¶

The lineage setting also decides where the datasets come from, and when BUSCO needs the network:

Setting	Where the datasets come from	Network
`auto`	every BUSCO job downloads its datasets into `eval/busco/busco_tmp/`, which is deleted when the job ends. Each sample downloads its own copy, because two BUSCO jobs downloading into one folder can corrupt it. The job downloads eukaryota_odb12, the placement files and the chosen lineage	in every BUSCO job, on the computer that runs it
a fixed lineage	rule `busco_data` downloads it once per output folder into `00.resources/busco/lineages/<lineage>/`, on the computer that runs Snakemake (`logs/busco_data.log`). Every BUSCO job reads it with `--offline` and writes nothing there, so the samples share one copy	once; the BUSCO jobs need none

With auto, the computer that runs BUSCO needs internet in every job

A fixed lineage needs internet once, on the computer that runs Snakemake, and the BUSCO jobs need none. This is how BUSCO runs on cluster nodes without internet (Network access).

With a fixed lineage, the JSON summary records offline: True and the path of the shared lineage folder, 00.resources/busco/lineages/<lineage>/. The .txt summary and full_table.tsv are the same as with a download per job.

The odb12 datasets of 2026-05-22 have these archive sizes:

Dataset	Archive size (MB)
eukaryota	29.1
fungi	125.8
ascomycota	310.5
sordariomycetes	311.7
hypocreales	274.6
basidiomycota	151.2
agaricomycetes	196.7
mucoromycota	132.1

With auto, a job therefore downloads tens to hundreds of MB, every time. This is what stays on disk afterwards:

In `eval/busco/`	After the run
the summaries	kept
`run_<lineage>/`: `full_table.tsv` (one line per gene with its status and location), `missing_busco_list.tsv`, `miniprot_output/`, BUSCO's own copy of the summary	kept
`auto_lineage/`, `logs/`, `tmp/`	kept
every `hmmer_output/` and `busco_sequences/` folder	deleted

The folder keeps hundreds of MB per sample.

FunFlux passes --opt-out-run-stats, so BUSCO 6.0.0 sends no usage statistics to busco-data.ezlab.org and makes no ipinfo.io lookup. Without the flag, BUSCO would send run details, results, a checksum of the input FASTA and the country of the machine's IP address. The flag changes no result.

Qualimap (read modes)¶

Qualimap shows how the sample's reads map back to the draft. Rule map_evaluation runs it in illumina, longread and hybrid mode:

unset DISPLAY
JAVA_TOOL_OPTIONS="-Djava.awt.headless=true" \
qualimap bamqc -bam contaminants/{sample}_map.bam --java-mem-size=<min(ram_gb, 64)>G \
  -nt <min(threads, 24)> -outdir eval/qualimap -outformat html

The BAM is the coverage track of the screen, the sample's reads aligned to the draft before decontamination (Decontamination). Every read-based mode gets a report on the reads that built the draft, hybrid a second one for the Illumina guide, and contigs mode none:

Mode	Rule	Reads in the BAM	Report in `02.assembly/{sample}/`
`illumina`	`map_evaluation`	the trimmed Illumina pairs (bowtie2)	`eval/qualimap/`
`longread`	`map_evaluation`	the filtered long reads (minimap2)	`eval/qualimap/`
`hybrid`	`map_evaluation`	the filtered long reads (minimap2)	`eval/qualimap/`
`hybrid`	`map_guide_evaluation`, the same command	the Illumina pairs mapped to the guide assembly (`guide/contaminants/{sample}_map.bam`)	`guide/eval/qualimap/`
`contigs`	none	its only BAM is the contigs aligned to themselves

The Java heap is the smaller of resources.ram_gb and 64 GB, and a changed ram_gb does not re-run the report (CPU cores and threads). The two settings in front of the command let Qualimap draw its plots on a server without a display. The Java runtime notes them with one Picked up JAVA_TOOL_OPTIONS line in the log, which is expected.

When Qualimap fails, the log ends with [FunFlux] map_evaluation: Qualimap exited with an error for sample S. ..., eval/qualimap/ stays in place, empty, and the MultiQC page has no mapping table for the sample.

The report gives the share of mapped reads, insert sizes of read pairs, mapping quality, GC content, error rate, the depth along the draft and, in genome_results.txt, the mean depth of every contig. The mitochondrion stands out by its depth, because a cell holds many copies of it. A contaminant often has a depth different from that of the isolate.

In the MultiQC report¶

MultiQC reads eval/quast/report.tsv, the BUSCO summaries and, in the read-based modes, eval/qualimap/ of every sample into 06.report/multiqc_report.html. In hybrid it also reads guide/eval/qualimap/. Each tool has one row per sample, with a second QUAST and Qualimap row for the guide in hybrid and a second BUSCO row with the automatic lineage:

Mode	QUAST rows	BUSCO rows	Qualimap rows
`contigs`	1	2 with the automatic lineage, 1 with a fixed one	none
`illumina`, `longread`	1	as above	1: `mapping QC \| <sample>`
`hybrid`	2	as above	2: `mapping QC \| <sample>` and `mapping Illumina QC \| <sample>`

The report also holds the contig summary table of every sample, with one row per contig of the draft assembly (Contig summary). Report describes the row names, the General Statistics table and the BUSCO folders that MultiQC skips.

Re-runs¶

Snakemake deletes eval/quast/, eval/busco/ and eval/qualimap/ before it runs the step again, so files you add to these folders are lost. These changes re-run the following steps:

Change	Re-runs
`contigs_final.fasta`, for example after a new decontamination policy or mitochondrion threshold	QUAST and BUSCO
the BAM, that is, a new draft or new reads	Qualimap
`busco.lineage`	BUSCO, the contig summary and MultiQC only, after `busco_data` has downloaded a newly named lineage

A new BUSCO lineage therefore re-runs only BUSCO and the two steps that read it, not the assembly or the annotation.

Output files¶

Everything these steps write is kept, except BUSCO's download folder, busco_tmp/, and the hmmer_output/ and busco_sequences/ folders inside eval/busco/.

Path	Kept	Read by
`02.assembly/{sample}/eval/quast/`	yes	MultiQC (`report.tsv`)
`02.assembly/{sample}/eval/busco/`	yes	MultiQC (the `short_summary.*.txt` files)
`02.assembly/{sample}/eval/busco/busco_tmp/` (`auto` only)	no	BUSCO (downloaded datasets)
`00.resources/busco/lineages/<lineage>/`, `00.resources/busco/file_versions.tsv` (fixed lineage only)	yes	BUSCO, in every later run into the output folder
`02.assembly/{sample}/eval/qualimap/`	yes (read-based modes)	MultiQC
`02.assembly/{sample}/guide/eval/qualimap/`	yes (`hybrid`)	MultiQC
`02.assembly/{sample}/eval/telomere_candidates.tsv`, `telomere_unit.txt`, `telomere_ends.tsv`	yes	the contig summary; you. The telomere repeat unit of the assembly, from tidk (Contig summary)
`02.assembly/{sample}/eval/contig_summary.tsv`, `contig_summary_mqc.tsv`	yes	you; MultiQC (`contig_summary_mqc.tsv`). Written after the annotation (Contig summary)
`06.report/multiqc_report.html`, `06.report/multiqc_data/`, `06.report/multiqc_config.yaml`	yes	you
`logs/genome_assembly_evaluation_{sample}.log`, `logs/completeness_{sample}.log`, `logs/busco_data.log` (fixed lineage), `logs/map_evaluation_{sample}.log`, `logs/map_guide_evaluation_{sample}.log`, `logs/telomere_units_{sample}.log`, `logs/contig_summary_{sample}.log`, `logs/multiqc.log`	yes	you

The full output tree is on Output files.