Skip to content

Latest commit

 

History

History
27 lines (21 loc) · 1.26 KB

README.md

File metadata and controls

27 lines (21 loc) · 1.26 KB
Raw

🧪 Fast ACSF Featurizer

High performance implementation of atom-centered symmetry functions featurizer. It featurizes ligand-receptor complex into a fixed size vector. It is heavly inspired by AQDnet featurizer.

Dependencies

Only dependencies are numpy and rdkit. You need to have C compiler installed to compile source code into shared library.

Usage

  1. Compile C source code using example command make compile.
  2. Import and define featurizer in Python.
  3. Pass it ligand atom coordinates and protein atom coordinates along with their respective atomic numbers.
from acsf.featurizer import Featurizer

featurizer = Featurizer(
    elements=['H', 'C', 'O', 'N'],
    angular_cutoff=8
)

⚡ Performance

Testing was done on i7-9700K processor with 32GB of 3200MHz RAM on Windows WSL (Ubuntu). Featurization of aa2ar complex from PDBBind dataset took on average 45ms per pose. Receptor has 4570 atoms, while ligand has 40. Default settings were used to generate feature vector with size 11583. It support also batch featurization and it is parallelized with OpenMP.

Num. of Poses Time
1 45ms
1000 6200ms