3D conformer generation
Generates one 3D starting geometry for a compound saved flat (e.g. a PubChem download), using either RDKit or Open Babel. Use before docking, or anywhere a 3D structure is needed and only a 2D one exists.
3D conformer generation
What it does
Takes one compound saved without 3D coordinates — the usual state of a PubChem download or a sketched structure — and produces one plausible 3D geometry, with explicit hydrogens added.
This is a prep step, not an analysis: the output is a starting point for whatever needs a 3D structure next (docking, most commonly), not a claim about the molecule's true lowest-energy shape.
Choosing an engine
Two engines are offered because they use genuinely different methods and can produce different starting geometries for the same molecule:
- RDKit (ETKDG), the default — builds the 3D structure from distance geometry (the known bond lengths and angles), then refines it with a molecular-mechanics force field (MMFF94). Deterministic here: the same molecule always produces the same conformer, so re-running with the same settings doesn't change the answer.
- Open Babel (force field) — generates the structure directly through force-field-based building rather than distance geometry first.
Neither is "more correct" than the other in general — they're two reasonable starting points, and if a downstream step behaves oddly with one, trying the other is a legitimate thing to do.
Reading the output
There's one file: a 3D SDF with explicit hydrogens. If a rule set couldn't fully optimize the geometry (RDKit's minimization didn't converge, or the force field had no parameters for part of the molecule), that's reported as a warning rather than a failure — the geometry is still returned, flagged as a rougher starting point than usual.
Input
One compound, saved as an SDF. If the file has more than one structure in it, only the first is used, and the result says how many records the file had.
Important limitations
One conformer, not an ensemble. This tool returns a single starting geometry, not a set of candidates to choose from or a search of conformer space.
Not a claim about the true 3D structure. Both engines produce a reasonable starting point; neither performs the kind of exhaustive conformational search a dedicated conformer-generation study would.
No protonation-state adjustment. The output reflects whatever protonation state the input file already encoded — this tool doesn't add or remove protons for a target pH.