neurosnap.algos.amber_relax module#

Amber-style relaxation / energy minimization utilities built on OpenMM.

This module loads a PDB, cleans it with PDBFixer (adds missing atoms/hydrogens, replaces nonstandard residues), removes waters and non-biopolymers, and then performs energy minimization (“relaxation”) using the Amber14 force field (amber14-all.xml with amber14/tip3p.xml). By default it uses HBond constraints, no PME (NoCutoff), and a Langevin integrator (300 K, 2 fs) for context initialization; only the minimizer is used to relax coordinates.

Requirements#

  • OpenMM (python package openmm) must be installed to use any function here.

  • pdbfixer is required for structure fixing.

  • NumPy is used for RMSD calculations.

  • Optional: CUDA-compatible GPU for faster context creation/minimization (set use_gpu=True, uses mixed precision and disables PME stream).

Key functions#

  • minimize(…): Run energy minimization on a PDB file. Accepts a tolerance in kcal/mol (clamped to [0.1, 10.0]), supports GPU/CPU selection, and returns a dict with rmsd (Å), initial_energy and final_energy (kJ/mol). B-factors from the input PDB are preserved and copied to the minimized output.

  • _compute_rmsd(…): Internal RMSD helper (nm in OpenMM units; returned as Å).

  • _load_bfactors_from_pdb(…) / _apply_bfactors_to_pdb(…): Internal helpers to snapshot and restore B-factors on ATOM/HETATM records.

Notes & caveats#

  • Uses NoCutoff for nonbonded interactions; adapt if you need PME/periodic systems. Hydrogens are constrained via HBonds.

  • Minimization stops when the maximum force falls below the specified tolerance (converted to kJ/mol) or when max_iterations is reached.

Example

>>> minimize("input.pdb", output_minimized_pdb="minimized.pdb",
...          tolerance=1.0, max_iterations=1000, use_gpu=True)
neurosnap.algos.amber_relax.minimize(pdb_file, output_minimized_pdb='minimized.pdb', max_iterations=1000, tolerance=1.0, use_gpu=True, properties={})[source]#

Loads a PDB file and performs energy minimization with a specified tolerance.

Parameters:

  • pdb_file (str) – Path to input PDB file.

  • output_minimized_pdb (str) – Path to output minimized PDB file.

  • max_iterations (int) – Number of minimization iterations.

  • tolerance (float) – Convergence tolerance in Kcal/mol (range: 0.1 - 10.0).

  • use_gpu (bool) – Whether to use GPU (CUDA). Defaults to True.

  • properties (dict[str, Any]) – Platform properties to provide to OpenMM if needed.

Returns:

Contains RMSD, initial energy, and final energy.

Return type:

dict