Expose Bemis–Murcko scaffolds and ETKDG/MMFF conformer generation

[ehartford]

Feature request

Could rdkit / rdkit-sys expose native bindings for Bemis–Murcko scaffold generation and 3D conformer generation/optimization?

The high-level crate already covers useful parsing and normalization operations, but these common workflows currently require downstream users to maintain their own CXX bridge.

Suggested surface

Bemis–Murcko:

• RDKit::MurckoDecompose(const ROMol &) from GraphMol/ChemTransforms
• A high-level helper that returns an owned molecule (or canonical isomeric SMILES)

3D conformers:

• RDKit::DGeomHelpers::ETKDGv3 / configurable EmbedParameters
• RDKit::DGeomHelpers::EmbedMolecule
• hydrogen addition/removal through MolOps::addHs and MolOps::removeHs
• MMFF parameter availability and optimization
• UFF optimization as a fallback
• molblock serialization with the generated conformer

A convenient high-level API might be split into composable operations, while a small options struct could cover the most important deterministic settings (random seed, small-ring torsions, random-coordinate retry, maximum iterations, and MMFF→UFF fallback).

Example workflow

1. Parse SMILES.
2. Generate a Murcko scaffold and canonicalize it.
3. Add hydrogens.
4. Embed with ETKDGv3 and a fixed seed.
5. Optimize with MMFF when parameters are available, otherwise UFF.
6. Remove hydrogens and serialize to a 3D molblock.

This would keep the entire workflow in-process and native for Rust applications, without each downstream crate recreating the same C++ wrappers and link-library list.

I’m happy to help test the API on macOS arm64 and Linux x86_64.