SMILES to PDB

What is SMILES to PDB?

SMILES to PDB is a small-molecule structure-generation and format-conversion workflow. It turns a SMILES molecular graph into a PDB file containing calculated three-dimensional atomic coordinates. SMILES compactly records connectivity, aromaticity, charge, and any stereochemistry that is written into the string. PDB is a coordinate-focused format that can be opened beside protein structures in molecular viewers such as PyMOL, ChimeraX, and VMD.

The generated structure is a starting conformer, useful for inspection, visualization, and handoff to software that specifically requests PDB. It is not an experimental structure, a predicted binding pose, or an ensemble of likely conformations. PDB also represents small-molecule chemistry less completely than SDF or MOL2, so SMILES to SDF is usually the better choice when a later step needs reliable bond-order information.

How to use SMILES to PDB online

ProteinIQ converts a SMILES string, named SMILES batch, uploaded text file, or PubChem record into one 3D PDB file per successful compound. Each molecule is embedded in three dimensions with RDKit, optionally hydrogenated and UFF-minimized, then returned with a curated Run log.

Inputs

Input	Accepted form	Details
SMILES	One valid SMILES per line	A batch accepts up to 10 entries. Aspirin, for example, is CC(=O)Oc1ccccc1C(=O)O.
Named SMILES	name<TAB>SMILES	The name identifies the result file. For example, aspirin\tCC(=O)Oc1ccccc1C(=O)O produces an aspirin-named PDB file.
File upload	.txt, .csv, .tsv, .smi, or .smiles	Uploaded text files can be up to 50 MB. Use one compound entry per line.
PubChem fetch	Compound name or PubChem CID	ProteinIQ retrieves the compound's SMILES representation before conversion.

Each entry is limited to 300 heavy atoms. Query atoms and very large ligand-like or peptide-like SMILES are rejected because standard 3D embedding can be impractical or unreliable at that scale. A protein or peptide sequence should be modeled with a structure-prediction method rather than converted through this small-molecule workflow.

The submitted string determines the chemical state being modeled. Protonation, tautomer, salt fragments, and stereochemistry are not selected or corrected automatically. For example, a neutral amine SMILES and its protonated form describe different inputs and can generate different structures.

Settings

Setting	Default	Effect
Add hydrogens	On	Adds explicit hydrogen atoms before coordinate generation. Keep this enabled when the output will be inspected as a complete molecular structure or passed to a workflow that needs hydrogen coordinates.
Optimize geometry	On	Runs up to 500 iterations of Universal Force Field (UFF) minimization after embedding. Turning it off preserves the initial embedded geometry.

Results

Result	Contents
Individual .pdb file	One generated conformer for each successful input, named from the supplied compound name or input label. The file contains PDB coordinate records for the generated molecule.
run.log	A curated run record with the input count, selected settings, conversion phases, successful and failed entry counts, and stable failure guidance.

Entries run independently. An invalid SMILES or a failed conformer embedding is recorded for that entry without discarding PDB files for other compounds in the batch. Unlike the SMILES to SDF workflow, SMILES to PDB does not create a combined batch structure file.

How SMILES to PDB generates a 3D structure

RDKit first parses the SMILES into a molecular graph. When Add hydrogens is enabled, explicit hydrogens are added before conformer generation. The workflow then runs RDKit's ETKDGv3 embedding method, a distance-geometry approach informed by experimental torsion preferences and chemical knowledge. A fixed random seed makes the primary embedding reproducible. If that attempt fails, RDKit makes a second attempt from random coordinates with the same method and seed.

With Optimize geometry enabled, the embedded conformer is relaxed with RDKit's Universal Force Field. The minimization can improve local bond lengths, angles, and nonbonded contacts, but it is not a quantum-chemical calculation and does not prove that the result is the lowest-energy conformer. If UFF has no parameters for an atom or does not converge within 500 iterations, that entry keeps its file and its row carries a warning. The resulting conformer is written directly as a PDB block.

Interpreting the PDB file

The PDB file supplies Cartesian coordinates, which is why it is convenient for placing a ligand next to a receptor in a viewer. It should be inspected before scientific interpretation: a plausible-looking conformer can still have the wrong protonation state, an unfavorable torsion, or a stereochemical choice inherited from the input SMILES.

PDB is not a complete small-molecule exchange format. Its standard coordinate records do not represent bond orders as directly as SDF or MOL2, and downstream programs may reconstruct ligand connectivity differently. The original SMILES and, when chemistry must survive the handoff, an SDF export should be retained as the authoritative ligand record.

For docking, a generated PDB is not automatically docking-ready. Docking engines often require their own charge model, atom typing, protonation decisions, and a format such as PDBQT. AutoDock Vina accepts SMILES and supported ligand structure files, then follows its docking-specific ligand preparation path.

When to use PDB, SDF, or MOL2

Goal	Preferred format or tool
View a generated ligand with a protein structure	PDB from SMILES to PDB
Preserve small-molecule bond orders and use a multi-record batch file	SMILES to SDF
Supply a program that explicitly requires Tripos atom and bond records	SMILES to MOL2
Convert among many chemistry and structure formats with format-specific options	Open Babel Format Converter

PDB is the practical option for coordinate visualization and PDB-only interfaces. SDF is generally safer for moving a ligand between cheminformatics and docking-preparation steps because it represents the molecular graph more explicitly.

Practical limits

• One conformer per compound: The workflow does not sample or rank a conformer ensemble. Flexible ligands and macrocycles may need dedicated conformer generation before using the structure for modeling decisions.
• Input state is retained: It does not enumerate stereoisomers, tautomers, or pH-dependent protonation states. The SMILES should be prepared for the intended chemistry before submission.
• Embedding can fail: Highly flexible, constrained, or unusual molecular graphs may not satisfy RDKit's 3D embedding constraints, even below the atom limit. The Run log records the failed-entry count while other batch entries continue.
• No target-aware modeling: The output has no receptor context. It should not be interpreted as a bound ligand conformation or binding-pose prediction.