---
name: distance-geometry-embedding
description: Use when when you have ionized adduct structures (in SMILES or MOL format) from an upstream ionization-state determination step and need to produce multiple low-energy 3D conformations for collision cross section prediction, metabolite annotation, or structure-property modeling.
license: CC-BY-4.0
metadata:
  edam_operation: http://edamontology.org/operation_0338
  edam_topics:
  - http://edamontology.org/topic_3314
  - http://edamontology.org/topic_0176
  tools:
  - Snakemake
  - RDKit
  - Dimorphite-DL
  - ASE-ANI
  techniques:
  - ion-mobility-MS
derived_from:
- doi: 10.1021/jasms.1c00315
  title: POMICS
evidence_spans:
- Snakemake workflow manager for predicting collisional cross sections
- This repository contains a Snakemake workflow manager for predicting collisional cross sections (CCS)
claims: []
provenance:
  collection: https://w3id.org/holobiomicslab/asb-skill/collection/metabolomics/v2
  assembled_by: scripts/collect_metabolomics_collection.py
  sources:
  - build: coll_pomics_cq
    doi: 10.1021/jasms.1c00315
    title: POMICS
  dedup_kept_from: coll_pomics_cq
schema_version: 0.2.0
attribution:
  generator: AgenticScienceBuilder
  original_doi: 10.1021/jasms.1c00315
  all_source_dois:
  - 10.1021/jasms.1c00315
  zenodo_doi: 10.5281/zenodo.20794027
  curators: []
  promoter: Louis-Félix Nothias
  sponsor: CNRS & Université Côte d'Azur
---

# distance-geometry-embedding

## Summary

Generate initial 3D molecular coordinates from 2D SMILES or MOL structures using RDKit's distance-geometry algorithm, producing a diverse ensemble of conformers that serve as starting geometries for subsequent force-field optimization and quantum filtering in CCS prediction workflows.

## When to use

When you have ionized adduct structures (in SMILES or MOL format) from an upstream ionization-state determination step and need to produce multiple low-energy 3D conformations for collision cross section prediction, metabolite annotation, or structure-property modeling. Use this skill when conformational diversity is required to adequately sample the potential energy surface before quantum refinement.

## When NOT to use

- Input molecules are already 3D structures with pre-optimized geometries suitable for immediate quantum calculation; distance-geometry embedding adds unnecessary computational overhead.
- Your workflow requires conformers at a specific energy level (e.g., QM-optimized); raw distance-geometry + MM minimization produces structures far from the QM ground state and will require subsequent re-optimization.
- Molecules contain non-standard elements or functional groups outside RDKit's MMFF94/UFF parametrization, causing minimization failures or invalid geometries.

## Inputs

- Ionized molecular structures (SMILES strings or MOL files)
- Integer parameter for number of conformers to generate per molecule (e.g., 50–500)

## Outputs

- 3D conformer ensemble (SDF or pickle format)
- Multiple conformations per molecule with hydrogens added and force-field-minimized geometries

## How to apply

Load ionized adduct structures output by Dimorphite-DL or similar ionization tools. Use RDKit's AllChem.EmbedMolecule function to generate multiple 3D conformers per molecule (typically 50–500 samples) via distance-geometry embedding, which builds 3D coordinates from inter-atomic distance constraints derived from the 2D connectivity graph. Add hydrogens to each structure before embedding. After embedding, minimize each conformer's geometry using RDKit's built-in force field (MMFF94 or UFF) to relax structures toward local minima. Export the complete conformer ensemble to SDF or pickle format, retaining all conformers for downstream filtering by ASE-ANI or other methods. The rationale is that distance-geometry produces diverse initial structures with reasonable bond and angle geometry without quantum cost, enabling efficient sampling of conformational space before expensive quantum calculations.

## Related tools

- **RDKit** (Performs distance-geometry embedding via AllChem.EmbedMolecule, adds hydrogens, and applies force-field geometry minimization (MMFF94/UFF) to each conformer) — https://www.rdkit.org
- **Dimorphite-DL** (Upstream tool that produces ionized adduct structures (SMILES/MOL format) that serve as input to distance-geometry embedding) — https://durrantlab.pitt.edu/dimorphite-dl
- **ASE-ANI** (Downstream conformation filtering tool that uses neural-network potentials to rank and filter conformers generated by distance-geometry embedding before quantum calculations) — https://github.com/isayev/ASE_ANI
- **Snakemake** (Workflow orchestration tool that automates and parallelizes distance-geometry embedding across multiple molecules on HPC systems) — https://github.com/DasSusanta/snakemake_ccs

## Examples

```
from rdkit import Chem; from rdkit.Chem import AllChem; mol = Chem.MolFromSmiles('[NH3+]C(C)C'); Chem.AddHs(mol); AllChem.EmbedMolecule(mol, numConfs=100); AllChem.MMFFOptimizeMoleculeConfs(mol); Chem.SDWriter('conformers.sdf').write(mol)
```

## Evaluation signals

- Conformer ensemble contains the expected number of 3D structures per input molecule (e.g., 50–500 conformers per molecule).
- All conformers are valid chemical structures with correct bond orders, valences, and hydrogen counts; no NaN or invalid coordinates in SDF/pickle output.
- Geometric validity: inter-atomic distances respect van der Waals constraints and typical bond/angle ranges (e.g., C–C bond ≈ 1.5 Å, C–H ≈ 1.1 Å).
- Force-field minimization convergence: reported energy per conformer decreases from initial embedding to final minimized state, indicating successful local geometry relaxation.
- Diversity of conformers: RMSD between conformers is > 0.5–1.0 Å (relative to heavy atoms), confirming that distance-geometry sampling explores distinct conformational regions rather than collapsing to a single structure.

## Limitations

- Distance-geometry embedding relies on RDKit's distance-bound matrix, which can fail for highly constrained or exotic ring systems, producing invalid or distorted 3D geometries.
- MMFF94 and UFF force fields are not parametrized for all elements; molecules containing transition metals, halogens beyond Cl, or other heavy atoms may fail minimization or produce unreliable geometries.
- Force-field minimization reaches only local minima; resulting conformers are not QM-optimized and may be far from ground state. Subsequent quantum filtering (ASE-ANI, QUICK) is essential to identify lower-energy structures.
- ASE-ANI repository is deprecated and no longer supported; users are directed to use TorchANI implementation instead, which may have different input/output formats.
- Conformer quality depends critically on the number of embedding attempts and force-field choice; empirically, 50–500 conformers may be insufficient for large, flexible molecules or those with multiple stable conformational isomers.

## Evidence

- [other] Use RDKit to add hydrogens and generate initial 3D coordinates via distance-geometry algorithm.: "Use RDKit to add hydrogens and generate initial 3D coordinates via distance-geometry algorithm"
- [other] Generate multiple conformers per structure using RDKit's AllChem.EmbedMolecule with specified number of conformers (e.g., 50–500 samples per molecule).: "Generate multiple conformers per structure using RDKit's AllChem.EmbedMolecule with specified number of conformers (e.g., 50–500 samples per molecule)"
- [other] Minimize each conformer's geometry using RDKit's built-in force field (MMFF94 or UFF) to relax structures toward local minima.: "Minimize each conformer's geometry using RDKit's built-in force field (MMFF94 or UFF) to relax structures toward local minima"
- [other] RDKit is used in the workflow as the conformation generation tool that operates on ionized adduct structures produced by Dimorphite-DL: "RDKit is used in the workflow as the conformation generation tool that operates on ionized adduct structures produced by Dimorphite-DL"
- [readme] This is a prototype interface for ANI-1x and ANI-1ccx neural network potentials for The Atomic Simulation Environment (ASE).: "This is a prototype interface for ANI-1x and ANI-1ccx neural network potentials for The Atomic Simulation Environment (ASE)"