This repository is intended to replicate only the colabfold_search functionality of the local ColabFold MSA generation pipeline, with an added protein database of 129944764 non-redundant proteins tailored for viruses, especially phages. Please see dataset_curation for more details on how it was constructed.

For the full Colabfold functionality and up-to-date changes (and frankly anything outside of the use case presented below), please go to the ColabFold repository.

This was used to create some MSAs used in phold's database.

• This will replicate the MSAs used in the enVhogs and PHROG singleton protein structures in Phold. You can use other, later versions of MMSeqs2 if you'd like.

• This assumes you have conda installed.

conda create -n colabfoldv_env python=3.12 mmseqs==15.6f452 pip
conda activate colabfoldv_env
git clone https://github.com/gbouras13/colabfoldv
cd colabfoldv
pip install -e .
pip install colabfold[alphafold]

• This will download the regular ColabFold uniref30_2302 and colabfold_envdb_202108 databases

mkdir -p colabfoldDBs
cd colabfoldDBs
bash ../setup_databases.sh

• The viral/phage database database is stored on Zenodo. It is 39GB, and therefore it may take some time to download.
• I would highly recommend using aria2c (as used in MMSeqs2 and Foldseek) to download, but you can use e.g. wget or curl

aria2c https://zenodo.org/records/15045387/files/viral_db.tar.xz
tar -xvf viral_db.tar.xz

• This will create MSAs without pairing (i.e. for monomers) for all proteins in example/NC_043029_aa.fasta using all three databases

THREADS=72
colabfold_search example/NC_043029_aa.fasta colabfoldDBs viral_db NC_043029_aa_msas --threads $THREADS