🚀 mini3di

A NumPy port of the foldseek code for encoding structures to 3di.

🗺️ Overview

foldseek is a method developed by van Kempen et al.[1] for the fast and accurate search of protein structures. In order to search proteins structures at a large scale, it first encodes the 3D structure into sequences over a structural alphabet, 3di, which captures tertiary amino acid interactions.

mini3di is a pure-Python package to encode 3D structures of proteins into the 3di alphabet, using the trained weights from the foldseek VQ-VAE model.

This library only depends on NumPy and is available for all modern Python versions (3.7+).

🔧 Installing

Install the mini3di package directly from PyPi which hosts universal wheels that can be installed with pip:

$ pip install mini3di

💡 Example

mini3di provides a single Encoder class, which expects the 3D coordinates of the Cα, Cβ, N and C atoms from each peptide residue. For residues without Cβ (Gly), simply write the coordinates as math.nan. Call the encode_atoms method to get a sequence of 3di states:

from math import nan
import mini3di

encoder = mini3di.Encoder()
states = encoder.encode_atoms(
    ca=[[32.9, 51.9, 28.8], [35.0, 51.9, 26.6], ...],
    cb=[[ nan,  nan,  nan], [35.3, 53.3, 26.4], ...],
    n=[ [32.1, 51.2, 29.8], [35.3, 51.5, 28.1], ...],
    c=[ [34.4, 51.7, 29.1], [36.1, 51.1, 25.8], ...],
)

The states returned as output will be a NumPy array of state indices. To turn it into a sequence, use the build_sequence method of the encoder:

sequence = encoder.build_sequence(states)
print(sequence)

The encoder can work directly with Biopython objects, if Biopython is available. A helper method encode_chain to extract the atom coordinates from a Bio.PDB.Chain and encoding them directly. For instance, to encode all the chains from a PDB file:

import pathlib

import mini3di
from Bio.PDB import PDBParser

encoder = mini3di.Encoder()
parser = PDBParser(QUIET=True)
struct = parser.get_structure("8crb", pathlib.Path("tests", "data", "8crb.pdb"))

for chain in struct.get_chains():
    states = encoder.encode_chain(chain)
    sequence = encoder.build_sequence(states)
    print(chain.get_id(), sequence)

💭 Feedback

⚠️ Issue Tracker

Found a bug? Have an enhancement request? Head over to the GitHub issue tracker if you need to report or ask something. If you are filing in on a bug, please include as much information as you can about the issue, and try to recreate the same bug in a simple, easily reproducible situation.

🏗️ Contributing

Contributions are more than welcome! See CONTRIBUTING.md for more details.

📋 Changelog

This project adheres to Semantic Versioning and provides a changelog in the Keep a Changelog format.

⚖️ License

This library is provided under the GNU General Public License v3.0. It includes some code ported from foldseek, which is licensed under the GNU General Public License v3.0 as well.

This project is in no way not affiliated, sponsored, or otherwise endorsed by the original foldseek authors. It was developed by Martin Larralde during his PhD project at the European Molecular Biology Laboratory in the Zeller team.

📚 References

• [1] Kempen, Michel van, Stephanie S. Kim, Charlotte Tumescheit, Milot Mirdita, Jeongjae Lee, Cameron L. M. Gilchrist, Johannes Söding, and Martin Steinegger. ‘Fast and Accurate Protein Structure Search with Foldseek’. Nature Biotechnology, 8 May 2023, 1–4. doi:10.1038/s41587-023-01773-0.