k-mers

Description and purpose of the project will come at a later point.

g++: through gcc installation

gzip / gunzip: usually pre-installed

python3: installed and part of path

optional: sqlite3: Debian/Ubuntu: apt install sqlite3; macOS: brew install sqlite3

Clone the repo

git clone https://github.com/cemiu/kmers.git && cd kmers

Two folders need to be populated:

• pdb (required)
• uniprotkb (optional)

The pdb folder has to contain experimental PDB files in the .ent.gz format.

Instructions for downloading can be found here:

https://www.rcsb.org/docs/programmatic-access/file-download-services

https://files.wwpdb.org/pub/pdb/data/structures/divided/pdb/

Alternatively a mirror can be found here: https://pycom.brunel.ac.uk/misc/pdb_2023-07-28.tar (42 GB)

Once downloaded they have to be placed in the pdb folder without being decompressed. It does not matter whether they are divided; e.g. pdb/file.ent.gz or pdb/<folder>/file.ent.gz.

Optionally, the uniprotkb folder can be populated with the uniprotkb fasta files. This is only needed, if the PDBs should be associated to a Protein. If this is not required, skip this step.

Files:

• uniprot_sprot.fasta.gz (400 MB after processing)
  • Use only Swiss-Prot, has the majority of PDB coverage
• Optionally, uniprot_trembl.fasta.gz (250 GB after processing)
  • Use TrEMBL to match more PDBs; might be useful for max. coverage

The latter might result in (slightly) more PDBs which can be associated to a Protein. The difference is expected to be trivial.

Place the files in the uniprotkb folder without decompressing them.

Once run.sh has been executed and the database uniprotkb/uniprot_sequences.db has been created, the files can be deleted.

To run the script, execute:

./run.sh

This will

• Compile C++ binaries, if they don't exist
• Ask whether to process the uniprotkb files
  • If yes, create the database uniprotkb/uniprot_sequences.db, if it doesn't exist
• Ask for k-mer size (default: k=12)
• Extracts 3d k-mer from the PDBs (into pdb_output folder)
• Extracts k-mer of length k into kmer.txt, along with frequency