q734738781/conformer-search-and-preopt

conformer-search-and-preopt

Overview

Use this skill when the task is a bounded molecular preoptimization episode before higher-level ORCA work.

Quick Start

1. If the input is a SMILES string, build a first 3D seed with create_molecule_from_smiles.
2. Use enumerate_molecular_conformers for an RDKit ETKDG seed ensemble.
3. Use remote_submission_batch with task_name="crest_run" when broad conformer-rotamer exploration is needed.
4. Use filter_conformer_ensemble to enforce one explicit energy/RMSD pruning policy.
5. Run remote_submission_batch with task_name="xtb_run" and params.mode="opt" on the accepted ensemble, then summarize it with analyze_xtb_results.
6. Use extract_optimized_molecules when the next stage needs a clean set of optimized XYZ files.

Allowed tools

• create_molecule_from_smiles
• enumerate_molecular_conformers
• filter_conformer_ensemble
• remote_submission
• remote_submission_batch
• analyze_xtb_results
• extract_optimized_molecules

Workflow

1. Keep one explicit ensemble contract

• Do not mix raw RDKit seeds, CREST ensembles, and xTB-optimized structures without a manifest.
• After every filtering step, emit the retained structure directory and the summary file path.

2. Separate exploration from refinement

• Use RDKit or CREST for ensemble generation.
• Use xTB for preoptimization, low-cost ranking, and quick Hessian checks.
• Do not skip the pruning step before higher-level work unless the task explicitly needs the full ensemble.

3. Hand off cleanly

• If the next stage is ORCA, pass a directory of optimized XYZ files rather than a mixed results tree.

Output Contract

Return:

• retained ensemble directory
• xTB summary path
• extracted optimized-molecule directory when generated