How to Use PULCHRA

PULCHRA is a fast protein chain restoration method for reconstructing all-atom protein models from reduced coordinate representations, especially C-alpha traces. It is useful when a structure prediction, coarse-grained model, decoy set, or simplified protein representation needs plausible backbone and side-chain atoms before visualization, scoring, docking preparation, or downstream structure cleanup.

On Neurosnap, researchers upload an Input Structure, choose a Reconstruction Mode, and optionally tune C-alpha optimization, geometry checks, hydrogen output, PDB-SG side-chain center-of-mass handling, and trajectory export. The service returns a rebuilt PDB structure plus a compact summary table showing how many atoms and residues were present before and after reconstruction.

PULCHRA reconstructs missing atomic detail from reduced protein coordinates using geometric chain restoration and fragment-library information. The method was designed for speed and robustness on C-alpha-only or partially reduced protein models, where the main goal is to recover a chemically interpretable all-atom representation without running an expensive simulation.

The core workflow rebuilds the protein backbone and side chains around the submitted C-alpha trace. Optional C-alpha optimization can adjust distorted traces within a user-defined maximum shift, while clash and chirality checks help avoid obvious local geometry problems. PDB-SG mode can use side-chain center-of-mass pseudoatoms when those are present, giving Pulchra more information than a C-alpha-only input.

Researchers should treat Pulchra outputs as reconstructed structural models, not experimental coordinates. The rebuilt model is often a practical starting point for inspection, filtering, atom-level scoring, or follow-up refinement with tools such as PDBFixer or AMBER Relaxation when force-field preparation or energy minimization is required.

• Reduced-model recovery: PULCHRA is designed for C-alpha and other simplified protein representations, filling a workflow gap between coarse models and atom-level tools.
• Configurable reconstruction: Researchers can rebuild full atoms, backbone atoms, or side chains and choose whether C-alpha positions should be optimized or preserved.
• Geometry-aware options: Clash handling, chirality checks, cis-proline detection, and hydrogen-bond optimization provide useful controls for model cleanup.
• Downstream-ready outputs: Neurosnap returns a rebuilt PDB and summary CSV so the reconstructed model can be visualized, downloaded, or passed into later scoring and refinement steps.