libmolgrid

libmolgrid converts three-dimensional molecular structures into multidimensional voxel grids and tensors for machine learning analyses in computational chemistry and bioinformatics.

Key Features:

• C++ core with Python bindings: Implemented in C++ with Python bindings for integration into Python-based computational workflows.
• Voxelization of molecular data: Transforms three-dimensional molecular structures into three-dimensional voxel grids.
• Tensor generation: Produces tensors from voxelized representations for consumption by deep learning frameworks.
• GPU acceleration: Supports GPU-accelerated computation for gridding and tensor creation.
• Neural network support: Facilitates use with convolutional and recurrent neural networks, including temporal and spatial recurrences over voxelized data.

Scientific Applications:

• Computational chemistry: Grid-based modeling of molecular interactions and molecular properties.
• Drug discovery: Development of predictive models for small-molecule activity using voxelized molecular representations.
• Protein–ligand binding studies: Analysis of protein–ligand interactions using voxel-based inputs to neural networks.

Methodology:

Transforms molecular structures into three-dimensional voxel grids, generates tensors from these voxelized data, employs GPU-accelerated computation for gridding and tensor creation, and supports temporal and spatial recurrences for convolutional and recurrent neural networks; implemented in C++ with Python bindings.