nf-core pipeline · metagenomics, protein-families
nf-core/proteinfamilies
Generates de novo protein families and updates existing models from amino acid sequence data.
What nf-core/proteinfamilies is
nf-core/proteinfamilies is a bioinformatics workflow designed to build protein families from amino acid sequences and update existing family profiles. Originally developed by Evangelos Karatzas, it groups polypeptide sequences into homologous clusters, constructs multiple sequence alignments, and generates profile Hidden Markov Models (HMMs).
Genomic and metagenomic studies frequently yield vast numbers of uncharacterised protein sequences that require systematic classification into functional or evolutionary groupings. The workflow automates sequence pre-processing, cluster identification, redundant family merging, and the integration of novel sequences into established family archives.
The pipeline also supports downstream structural and functional research by optionally exporting samplesheets directly formatted for execution in nf-core/proteinfold and nf-core/proteinannotator.
What the pipeline does
Input amino acid sequences in FASTA format are evaluated for quality metrics using SeqFu, followed by pre-processing in SeqKit to strip gaps, convert residues to upper case, filter by sequence length, replace special characters, and remove exact duplicates. For de novo family creation, sequences are clustered with MMseqs2 and aligned using either FAMSA or MAFFT, with optional gap trimming via ClipKIT. HMMER then constructs family HMMs and searches for additional sequence members. Alternatively, the iterative mgnifam algorithm can replace standard alignment and profile generation steps, refining alignments and models across up to three cycles or until model convergence.
To refine family boundaries, HMMER compares representative sequences against family models to merge similar families or remove redundancy. In-family redundant sequences can optionally be culled using strict MMseqs2 clustering, and remaining Stockholm-formatted alignments (.sto) are reformatted to FASTA (.fas) using HH-suite3. CMAPLE can optionally infer maximum parsimonious likelihood estimation phylogenetic trees from the resulting alignments, and MultiQC generates diagnostic visualisations of family size distributions and representative sequence lengths. When updating existing collections, input sequences are queried against provided family models; matching sequences are appended to their respective families and realigned, whilst non-matching sequences enter the de novo generation branch.
Key facts
- Pipeline type
- Nextflow workflow
- Latest release
- 2.5.0
- Primary inputs
- Amino acid FASTA files (uncompressed or gzip-compressed)
- Optional inputs
- Existing family HMM and MSA archives (.tar.gz)
- Clustering tools
- MMseqs2
- Alignment tools
- FAMSA, MAFFT, ClipKIT
- Profile model tool
- HMMER
- Phylogeny tool
- CMAPLE
- Reporting
- MultiQC, SeqFu
What you can use it for
- De novo family construction
- Cluster unannotated amino acid sequences from metagenomic or genomic datasets to build novel multiple sequence alignments and profile HMMs.
- Family database updating
- Screen newly sequenced proteins against existing family HMMs to update alignments and profile models without recomputing unchanged families.
- Iterative model refinement
- Execute the mgnifam algorithm to cycle through alignment, gap trimming, and model building up to three times or until convergence is reached.
- Phylogenetic reconstruction
- Derive maximum parsimonious likelihood estimation trees for completed family alignments using CMAPLE.
- Downstream workflow preparation
- Automatically generate samplesheets to transfer resulting families into nf-core/proteinfold or nf-core/proteinannotator.
Run it here
The research agent runs nf-core/proteinfamilies 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 both de novo family generation and incremental updates to existing family collections in a single unified execution.
- Provides a choice between FAMSA and MAFFT for multiple sequence alignment, coupled with optional ClipKIT alignment trimming.
- Features the mgnifam iterative refinement mode that optimises family alignments and profile models across up to three rounds.
- Includes automated redundancy reduction steps to merge closely related families and remove within-family sequence duplicates.
Watch out for
- Requires that input files for existing HMM and MSA archives share identical base filenames in equal numbers.
- Operates exclusively on amino acid sequences, requiring nucleotide coding sequences to be translated prior to submission.
- Phylogenetic tree inference is restricted to the maximum parsimonious likelihood estimation method implemented in CMAPLE.
Questions
Can I run nf-core/proteinfamilies without my own cluster?
The research agent runs nf-core/proteinfamilies 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/proteinfamilies under?
MIT, like every nf-core pipeline.
Can the pipeline process novel sequences and update existing families at the same time?
Yes. When existing HMM and MSA archives are supplied in the samplesheet, sequences matching existing models are incorporated into those families, while sequences without hits are routed into the de novo family generation workflow.
What is the mgnifam algorithm within this pipeline?
The mgnifam routine is an iterative alternative to standard alignment and model construction that repeats alignment, trimming, and HMM generation for up to three rounds or until the family profile converges.
What inputs are required to update existing protein families?
Users must provide the new amino acid FASTA file alongside tarball archives (.tar.gz) of existing family HMM and MSA directories, where corresponding files share matching base names.
How does the pipeline handle sequence quality and formatting issues?
SeqKit is used to filter sequences by length, convert residues to upper case, remove gaps and duplicate sequences, and replace special characters before clustering commences.
How to cite it
Evangelos Karatzas, Martin Beracochea, Fotis A. Baltoumas, Eleni Aplakidou, Lorna Richardson, James A. Fellows Yates, Daniel Lundin, nf-core community, Aydin Buluç, Nikos C. Kyrpides, Ilias Georgakopoulos-Soares, Georgios A. Pavlopoulos & Robert D. Finn. nf-core/proteinfamilies: A scalable pipeline for the generation of protein families. GigaScience. 2026 Jan. doi: 10.1093/gigascience/giag009.
Written from nf-core/proteinfamilies's own README. Statements citing a figure that does not appear there are removed before publication. Licence and running details come from our registry.