AGL-Score (Accelerated)

Fast, C++ accelerated calculation of Multiscale Weighted Colored Subgraphs (MWCGS) for protein-ligand complexes.

This package calculates algebraic graph features (AGL-Score) for molecular binding interfaces. It uses a hybrid C++/Python architecture (via Pybind11 and Eigen) to achieve orders of magnitude speedup over pure Python implementations.

---

🛠 Prerequisites

Before installing the Python package, you must have the following C++ libraries installed on your system.

1. OpenMP (Required for Parallelism)

• macOS:

  brew install libomp

• Ubuntu/Debian:

  sudo apt-get install libomp-dev

2. Chemfiles (Required for PDB/Mol2 Parsing)

The easiest way to install Chemfiles is via Conda/Mamba, which places the headers and libraries where the build system can find them.

conda install -c conda-forge chemfiles
conda install xgboost scikit-learn numpy pandas