nf-core pipeline · eukaryotes, metagenomics, metatranscriptomics
nf-core/metatdenovo
Assemble and annotate metatranscriptomic and metagenomic sequencing data for prokaryotes, eukaryotes and viruses.
What nf-core/metatdenovo is
nf-core/metatdenovo is a bioinformatics pipeline designed for the de novo assembly and annotation of metatranscriptomic and metagenomic sequencing data. It supports analyses across prokaryotes, eukaryotes and viruses, addressing the complexity of reconstructing diverse community transcripts and genomes from mixed environmental or biological samples. Danilo Di Leo, Emelie Nilsson and Daniel Lundin originally developed the pipeline within the nf-core framework.
The workflow unifies read quality assessment, assembly, open reading frame identification, gene quantification and combined functional and taxonomic annotations into an automated procedure. Output tables are provided in standardised formats suitable for statistical exploration and downstream biological interpretation in tools such as R or Python.
What the pipeline does
The analysis begins with read-level quality control via FastQC and MultiQC, followed by adapter and quality trimming using Trim Galore!. Optional read filtering and sequence depth normalisation can be performed with BBDuk and BBNorm, respectively. Trimmed paired-end reads are merged using Seqtk before de novo assembly, which is carried out using either SPAdes (with rnaSPAdes as the default setting) or MEGAHIT. Assembly metrics are subsequently evaluated using QUAST.
Open reading frames (ORFs) are predicted using one or more configurable tools: TransDecoder or MetaEuk for eukaryotes, Prodigal or Prokka for prokaryotes, or custom user-supplied ORFs. When MetaEuk is used, reference protein databases are automatically retrieved if not supplied. Detected ORFs on the same contig are merged into loci and clustered across contigs using MMseqs2. Gene expression or abundance is quantified by indexing the assembly and mapping cleaned reads with BBMap, followed by count generation using FeatureCounts. Functional characterisation is achieved through Prokka, eggNOG-mapper, KofamScan, dbCAN for CAZyme annotation, and HMMER against user-selected profile HMMs, while taxonomic assignments are derived using EUKulele and DIAMOND.
Key facts
- Pipeline engine
- Nextflow
- Latest release
- 1.4.0 (2026-06-26)
- Original authors
- Danilo Di Leo, Emelie Nilsson, Daniel Lundin
- Input format
- CSV samplesheet specifying raw FASTQ files (.fq, .fastq, .fq.gz, .fastq.gz)
- Assemblers supported
- SPAdes (rnaSPAdes default), MEGAHIT
- ORF callers supported
- TransDecoder, MetaEuk, Prokka, Prodigal, or user-supplied
- Functional annotation tools
- eggNOG-mapper, KofamScan, dbCAN, HMMER, Prokka
- Taxonomic annotation tools
- EUKulele, DIAMOND
- Primary output formats
- Gzipped TSV and optional Parquet summary tables
What you can use it for
- Eukaryotic and prokaryotic metatranscriptome assembly
- Assemble mixed microbial community transcriptomes de novo using rnaSPAdes or MEGAHIT, followed by splice-aware eukaryotic ORF calling with MetaEuk or prokaryotic gene calling with Prodigal.
- Metagenomic gene quantification
- Map trimmed reads back to assembled contigs using BBMap to generate gene-level feature counts for relative abundance estimation across conditions.
- Carbohydrate-active enzyme profiling
- Screen community ORFs using dbCAN to detect and classify CAZyme families involved in complex carbohydrate degradation.
- Taxonomic and functional annotation of uncultured viruses and microbes
- Classify assembled genes taxonomically with EUKulele and DIAMOND, alongside pathway and domain assignment through eggNOG-mapper, KofamScan and HMMER profile searches.
Run it here
The research agent runs nf-core/metatdenovo with Nextflow on its own machine — up to 32 cores, 256 GiB of memory and 24 hours per run, which covers the test profile and real single-sample runs. It builds the samplesheet, checks it against the pipeline's schema and reads the results back. When a run is too big for one machine, it gives you the exact command for your own cluster instead.
Strengths and limits
Good at
- Supports multiple dedicated ORF calling algorithms suited to both spliced eukaryotic genes and prokaryotic features within the same unified pipeline.
- Combines multiple functional annotation tools including dbCAN, KofamScan, eggNOG-mapper and HMMER in a single run.
- Generates standardised, consistent summary tables in TSV and optional Parquet formats optimised for direct analysis in R and Python.
- Consolidates overlapping ORF calls into discrete loci and clusters non-redundant proteins across contigs using MMseqs2.
Watch out for
- Requires paired-end and single-end reads to be kept separate per sample name, as the samplesheet cannot mix library types for the same sample.
- Computationally demanding when deploying comprehensive functional databases and assemblers, though resource scaling can be adjusted via MEGAHIT.
Questions
Can I run nf-core/metatdenovo without my own cluster?
The research agent runs nf-core/metatdenovo with Nextflow on its own machine — up to 32 cores, 256 GiB of memory and 24 hours per run, which covers the test profile and real single-sample runs. It builds the samplesheet, checks it against the pipeline's schema and reads the results back. When a run is too big for one machine, it gives you the exact command for your own cluster instead.
What licence is nf-core/metatdenovo under?
MIT, like every nf-core pipeline.
Can I assemble both single-end and paired-end sequencing libraries?
Yes. A samplesheet can include both single-end and paired-end sequencing files, provided that individual samples are exclusively single-end or paired-end, rather than a mix for the same sample identifier.
Which de novo assembler should I choose?
SPAdes (with rnaSPAdes as default) is recommended for both prokaryotic and eukaryotic assemblies, whereas MEGAHIT is provided as an alternative that requires fewer computational resources.
How are eukaryotic splice junctions handled during gene calling?
MetaEuk can be selected as the ORF caller for eukaryotic data; it is splice-aware and aligns assemblies against reference protein databases, which are fetched automatically if not provided.
How are gene expression levels quantified from the assembled contigs?
Cleaned reads are aligned back to an index of the assembly using BBMap, after which FeatureCounts aggregates read counts across identified genes into a unified TSV output.
In what format are the final annotation tables produced?
The pipeline outputs standardised tables in the summary_tables folder as gzipped TSV files, and can optionally export matching Parquet files if the save_parquet parameter is specified.
How to cite it
Di Leo, Danilo, Emelie Nilsson, Arianna Krinos, Jarone Pinhassi, and Daniel Lundin. PeerJ 13 (December 2025): e20328. https://doi.org/10.7717/peerj.20328.
Written from nf-core/metatdenovo's own README. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.