Heme

Heme is a Rust library for working with proteins

Features

Read and write common formats for protein data

Heme can read PDB files and construct a Pose object (named after the similar concept from Rosetta).

use io::parse_pdb; 
use conformation::Pose; 

let atoms = parse_pdb(&contents);
let pose = Pose::from_atoms(atoms);

The pose object has some useful functions, aside from being a container

• Pose object representing a protein structure
• Atom object that represents a single atom
• both Pose and Atom have methods that are useful for calculating structural features, such as the distance between two atoms, or the residues within some cutoff of a provided atom

Fast featurization for graph-structured generative models

Heme is a fast way to featurize protein structures for use with machine learning models

heme featurize path/to/pdbs --output-format json --output-file dataset.json 

Molecular modeling

Heme includes an implementation of force-field based molecular modeling in the Rust language. It's a work in progress that's helping me learn Rust. I hope that it will eventually be useful, but right now it's not. (For good implementations of force-field based molecular modeling, see the OpenMM or Lumol projects.)

Roadmap

• Heme 0.4.0 (planned)
  • implement fast featurization for structure transformer
  • featurization for ProteinMPNN/LigandMPNN compatibility
• Heme 0.3.2
  • handle ligands, metal ions, etc
• Heme 0.2.0
  • basic geometric operations
  • basic force field implementation
• Heme 0.1.0
  • basic PDB parsing
  • Atom and Record objects