PDB to SDF Converter (ligand extractor)

In short: Paste a PDB file or open one. Pick the ligand (HETATM group), and get an SDF file with its 3D coordinates, bonds and bond orders. It is free, runs in your web browser and needs no download.

Works for a small-molecule PDB and for a protein–ligand complex. Bonds come from CONECT records or from atom distances. Bond orders come from duplicated CONECT entries, or are estimated from the geometry.

How to convert PDB to SDF

1. Paste the PDB text or press Open .pdb file.
2. Choose the ligand in Which atoms to convert. For a file that holds one small molecule, use Whole file.
3. Keep Estimate bond orders ticked if the file has no bond orders (most PDB entries).
4. Press Convert to SDF. Check the atom count, the bond source and the SMILES under the button.
5. Press Download .sdf or Copy.

PDB and SDF: what changes

PDB (Protein Data Bank format) stores one atom per line with fixed columns: atom name, residue, chain and x, y, z in ångström. Ligands, ions and water are HETATM records. SDF (structure-data file) stores a MOL block per molecule: atoms, bonds with bond orders and charges, then optional data fields. Docking, cheminformatics and drawing programs read SDF.

How bonds are found

A PDB file stores atoms, not bonds. The converter uses CONECT records when the file has them. Atoms with no CONECT entry are bonded by distance: two atoms are bonded when they are closer than the sum of their covalent radii plus 0.45 Å. Hydrogen keeps one bond, carbon at most four.

How bond orders are found

Some programs (Open Babel, RDKit, PyMOL) write a double bond as two CONECT entries. When the file has such duplicates, they give the bond orders directly. Files from the Protein Data Bank do not have them. Then, with the estimate box ticked, the tool works out bond orders from the geometry: it decides whether each atom is sp, sp2 or sp3 from bond angles and bond lengths, finds flat 5- and 6-membered rings, gives aromatic rings alternating double bonds, and sets C=O, C=C, C=N and similar double bonds where the bond is short. This works well for well-resolved ligands, but it is an estimate. Always check the result.

What the SDF holds

One record: the MOL block (V2000, or V3000 above 999 atoms) with the PDB coordinates, formal charges from PDB columns 79–80, and data fields for the source group, the bond source, the bond-order method and the SMILES.

Limitations

• Bond orders estimated from geometry can be wrong for poorly resolved ligands, metal complexes, unusual tautomers and charged groups (for example a carboxylate is written as C=O plus O–H).
• No hydrogens are added and no protonation state is chosen.
• Only the first MODEL and the first alternate location (usually A) are read.
• Bonds between the ligand and the protein (covalent ligands) or metal ions are not written.
• Formal charges are only kept when the PDB file has them in columns 79–80.