Snakemake pipeline to cluster proteomes in two steps with MMseqs2, MAFFT, HHsuite and MCL:
- Proteins are clustered into subfamilies in a first round using MMseq2.
- For each subfamily, a MSA is generated using MAFTT.
- MSA profiles are compared all-vs-all using hhblits.
- Subfamilies are clustered into families using MCL based on probability and coverage obtained with hhblits.
Pipeline is written to run as an Snakemake workflow. After cloning the repo,
create a conda environment spc containing Snakemake:
# clone the repo
$ git clone https://github.com/dcarrillox/two-steps-clustering.git
$ cd two-steps-clustering
# create minimum environment for execution
$ conda create -n spc --file=conda-linux-64.lock
# activate environment
$ conda activate spc
# test Snakemake
$ (spc) snakemake --version
6.6.1
Pipeline's input is a FASTA file with all the proteins to be clustered. The path
to this file needs to be provided in the config file config/config.yaml under the
proteome section. Parameters for execution can be tuned in the config file as
well.
To run the pipeline:
# adjust number of threads with the -j argument
$ snakemake -j 16 --use-conda --conda-frontend mamba
Final results with the proteins clustered into subfamilies and families should be
under results/protein_clustering_results.tsv.