Read the paper — bioRxiv · DOI: 10.64898/2026.10.08.757395
Install · Quickstart · Documentation · Examples · Paper code
FastaLake uses de novo peptide predictions to select compact protein databases for metaproteomics. It retrieves complete proteins from a large sequence reservoir, selects a database for each mass spectrometry acquisition, and runs Sage to identify peptides. Study-wide protein grouping and quantification turn those results into a matrix ready for analysis.
- Select complete proteins from public sequence catalogues using exact peptide lookup.
- Search and quantify a study from one acquisition manifest.
- Add matched metagenomic or metatranscriptomic sequences when available.
- Process large reservoirs in chunks when memory is limited.
- Inspect peptide evidence, protein-group quantities and QC plots.
| Input | Supported route |
|---|---|
| Spectra | One mzML or mzML.gz file per DDA acquisition |
| De novo predictions | AlphaNovo CSV or generic sequence,score CSV, optionally gzipped |
| Protein sequences | Prepared protein FASTA reservoir |
| Matched molecular data | Optional specimen-matched DNA/RNA sequences and abundance tables |
RAW conversion and de novo inference happen before FastaLake. See accepted inputs and preparing a study.
| Route | Best for |
|---|---|
| Docker | A first run with the tools and examples already bundled |
| Source | Python development or a Linux cluster |
| Apptainer / Slurm | Running a study on a research cluster |
With Docker installed:
docker pull --platform linux/amd64 ghcr.io/mannlabs/fasta-lake:v1.1.0rc2The image includes Python, compiled Rust tools, Sage, directLFQ, AlphaPeptTools
and example inputs. It targets Linux x86_64; Apple Silicon uses Docker's x86_64
emulation. The published release is v1.1.0rc2; development changes on main
are available through the source or local Docker build instructions.
Run two small, real CAMPI acquisition windows and write the results locally:
mkdir -p fastalake-results
docker run --rm --platform linux/amd64 --network none \
--memory 4g --memory-swap 4g --cpus 2 \
--user "$(id -u):$(id -g)" -v "$PWD/fastalake-results:/work" \
ghcr.io/mannlabs/fasta-lake:v1.1.0rc2 \
fastalake-example campi --out /work/campi_01 --threads 2Open these files in fastalake-results/campi_01/:
| File | Result |
|---|---|
study/qc/QC_overview.png |
Quantities, group counts and completeness across acquisitions |
study/study_group_matrix.tsv |
Protein-group quantities |
study/peptide_evidence.tsv.gz |
Peptides supporting each group |
workflow/samples/S01/inference/S01_razor.fasta |
The selected protein database |
The example checks its outputs and saves VALIDATION.json. Use a new output
folder for each run. Continue with your first study
or reading the results.
| Example | What you learn |
|---|---|
| CAMPI | Selection, searching, grouping and quantification |
| Chunked lookup | Processing a reservoir within a memory budget |
| Matched DNA and RNA | Adding molecular evidence to the selected database |
| Community analysis | Exploring taxonomy and peptide–protein links |
Browse the tutorials, follow the visual walkthrough, or use an assistant task file to work through a study.
Treit et al. (2026). FastaLake selects compact metaproteomic databases from large sequence reservoirs by de novo peptide lookup. bioRxiv. Please also cite the software version used; see CITATION.cff. The manuscript analyses and source tables are maintained in fastalake-paper.
FastaLake uses Sage, directLFQ and AlphaPeptTools. Please cite these projects when using their results.
Report a reproducible problem in Issues. For setup, tests and pull requests, see Contributing. Code is under the MIT licence; bundled data and dependencies retain their own terms.