A no-code tool that turns a cofactor or ligand into a docking-ready 3D structure (SDF + PDB).
Type a name such as FAD, NADPH or hexadecane, paste a SMILES, or upload a 2D file. The tool builds and energy-minimises a 3D structure with RDKit and saves it for PyRx, AutoDock Vina or any other docking program. It runs in your web browser, on your own computer, and needs no coding.
| Step | What to do | How often |
|---|---|---|
| 1 | Install Python from python.org/downloads. |
once |
| 2 | ⬇️ Download the tool (ZIP), right-click the ZIP and choose Extract All, then open the folder and double-click Install first time.bat. Wait until it says Done! |
once |
| 3 | Double-click Start Cofactor 3D.bat. The tool opens in your web browser. |
every time |
Keep the black window open while you use the tool, and close it when you have finished. Your 3D files are saved in Documents\Ligands_3D.
Something not working? See Troubleshooting.
Building 3D ligands from SMILES with a script is quick, but it is easy to get wrong without noticing:
- Missing stereochemistry. If a SMILES has no
@/@@, RDKit picks the stereocentres at random. For cofactors like FAD or NAD(P)H you can end up docking an L-sugar or a mixed isomer, and the formula still looks correct. This tool stops and warns you when stereocentres are undefined. - Wrong redox form or tautomer. For example, a hand-written "FADH2" SMILES can be a 4a,5-dihydro tautomer instead of the true 1,5-dihydroflavin. Fetching the structure from PubChem by name avoids this, and the tool shows the PubChem CID so you can check it.
- One random conformer. The tool builds several conformers and keeps the lowest-energy one, with a fixed random seed so every run is reproducible.
- Poor record keeping. Every run is logged (SMILES, seed, force field, energy) in a CSV file, so you can write your Methods section later.
- Three input options: search by name (PubChem), paste SMILES, or upload
.sdf/.mol/.mol2/.smifiles - Quick-pick buttons for common cofactors (FAD, FADH2, FMN, FMNH2, NAD⁺, NADH, NADP⁺, NADPH, CoA) and n-alkanes (n-C16, n-C22, n-C28)
- Guard against undefined stereochemistry, which you can override for achiral molecules such as alkanes
- Removal of salts and water: only the largest fragment is kept
- Interactive 3D preview in the browser (3Dmol.js)
- Output as SDF (keeps bond orders; recommended for PyRx) and PDB (with a residue code you choose)
- Optional multi-conformer SDF for ensemble docking
generation_log.csvrecording every structure you make
Open a Terminal inside the downloaded folder and run:
pip install -r requirements.txt
python cofactor3d_app.pyThe tool opens in your browser. If it doesn't, go to http://127.0.0.1:8765.
- Choose your molecule. Search by name, paste a SMILES or upload a file. When searching by name, open the PubChem link and check that it is the form you want, for example oxidised vs reduced.
- Name it, and optionally set a 3-letter PDB residue code (e.g.
FAD,FDA,NDP). - Click Make 3D structure. Check the 3D preview, then use the SDF file for docking.
Files are saved to Documents\Ligands_3D by default. To change this, create a settings.json file next to cofactor3d_app.py:
{"output_folder": "D:\\My project\\Docking\\Ligands_3D"}- The input is parsed with RDKit. Only the largest fragment is kept, and undefined tetrahedral or double-bond stereocentres are detected; phosphorus and sulfur centres are ignored.
- Hydrogens are added, and n conformers (default 10) are embedded with ETKDGv3 (random seed 42 by default).
- Each conformer is optimised with MMFF94, or UFF if MMFF94 parameters are missing.
- The lowest-energy conformer is written as SDF and PDB.
Method references:
- ETKDG: Riniker S, Landrum GA (2015). J Chem Inf Model 55:2562–2574.
- ETKDGv3: Wang S, Witek J, Landrum GA, Riniker S (2020). J Chem Inf Model 60:2044–2058.
- MMFF94: Halgren TA (1996). J Comput Chem 17:490–519.
Example Methods sentence (please adapt it to your settings):
Three-dimensional structures of the cofactors were obtained from PubChem isomeric SMILES and generated with RDKit (version X) using the ETKDGv3 algorithm (10 conformers, random seed 42), followed by MMFF94 geometry optimisation. The lowest-energy conformer was used for docking.
- Testing: tested on Windows 11 with Python 3.14 and RDKit 2026.03. Mac and Linux should work but have not been tested yet; please report any problems.
- Privacy: the tool runs only on your own computer (
127.0.0.1) and only accepts requests from its own page. Nothing you draw or upload is sent anywhere, except a name search, which sends the name you type to PubChem. - Metal-containing cofactors (heme, Fe–S clusters, di-iron centres) cannot be built reliably. Take them from an experimental structure in the PDB.
- Protonation state: PubChem SMILES are usually neutral (e.g. phosphates drawn as –OH). The tool does not protonate for a particular pH, so prepare the charge state in your docking software if you need to, and report it.
- A single minimised gas-phase conformer is a starting point. Docking programs such as Vina still treat rotatable bonds as flexible.
- The name search needs internet access to PubChem; the 3D preview needs internet to load 3Dmol.js. Building and saving work offline with SMILES or files.
| Problem | Solution |
|---|---|
| "Python was not found" | Install Python (Quick start step 1) and make sure "Add python.exe to PATH" is ticked. If you already installed it without ticking, run the installer again, choose Modify, and tick it. |
| The black window says "No module named rdkit" | Run Install first time.bat first, and check that it ends with Done! |
| Windows shows "Windows protected your PC" | Click More info, then Run anyway. The .bat files are plain text, and you can open them in Notepad to see exactly what they do. |
| The browser did not open | Leave the black window open, and type http://127.0.0.1:8765 in your browser's address bar. |
| "Lost contact with the tool" on the web page | The black window was closed. Double-click Start Cofactor 3D.bat again. |
| "Could not reach PubChem" | Check your internet connection and try again in a minute (PubChem is sometimes busy), or paste the SMILES instead. |
| The tool refuses because of undefined stereocentres | This protects you from docking a random isomer. Use Search by name, or copy the SMILES from PubChem. For molecules without stereocentres (e.g. alkanes), tick Advanced settings → Allow undefined stereo. |
| No 3D preview appears | The preview needs internet. Your SDF/PDB files are still created. |
| Something else | The error is saved in error_log.txt in the tool folder. Please open an issue and attach it. |
-
PubChem (NCBI/NLM). The name search uses the public PUG-REST web service. It retrieves only the CID, title, molecular formula and SMILES of the one compound you search for; no PubChem data is stored in or redistributed with this repository. The tool follows PubChem's usage policy: at most 2 requests per second (the limit is 5), backing off when PubChem is busy, caching repeat searches, and sending a User-Agent that identifies this tool. This tool is independent and is not affiliated with or endorsed by NCBI/NLM; see the NCBI policies and disclaimer. If you use structures obtained through this tool, please cite PubChem as it requests:
- Kim S, Chen J, Cheng T, et al. PubChem 2025 update. Nucleic Acids Res. 2025;53(D1):D1516–D1525. doi:10.1093/nar/gkae1059
- Kim S, Thiessen PA, Cheng T, Yu B, Bolton EE. An update on PUG-REST: RESTful interface for programmatic access to PubChem. Nucleic Acids Res. 2018;46(W1):W563–W570. doi:10.1093/nar/gky294
In your Methods, also give the PubChem CID of each compound you used; the tool shows it and writes it to
generation_log.csv. -
RDKit, the open-source cheminformatics toolkit (BSD licence): https://www.rdkit.org
-
3Dmol.js, for the 3D viewer (BSD licence): Rego N, Koes D. Bioinformatics 2015;31(8):1322–1324.
If this tool helps your research, please cite it using the "Cite this repository" button on this page (see CITATION.cff). Please also cite RDKit.
Vidumini Alwis, MPhil researcher, Department of Zoology and Environment Sciences, University of Colombo, Sri Lanka · ORCID 0009-0004-6155-2295
Developed with the help of Claude (Anthropic) as an AI coding assistant.
MIT: free to use, modify and share, with attribution.