Molecular Property Calculator

What this calculates

Everything comes from RDKit compiled to WebAssembly, so the numbers are the same ones you would get from a Python script using the same toolkit.

• Identifiers: canonical SMILES, SMARTS, InChI, InChIKey, molecular formula, formal charge, MOL file.
• Physicochemical properties: average and monoisotopic mass, cLogP and molar refractivity (Wildman–Crippen), topological polar surface area (Ertl), Labute ASA, Fsp3.
• Counts: heavy atoms, heteroatoms, hydrogen bond donors and acceptors, rotatable bonds, amide bonds, rings, aromatic and saturated rings, heterocycles, spiro and bridgehead atoms, stereocentres including unspecified ones.
• Rule sets: Lipinski, Veber, Ghose, Egan, Muegge, rule of three and CNS criteria, each showing which specific criterion failed rather than a bare pass or fail.
• Ligand efficiency: add an IC50 or Kd and get LE, LLE, LELP, BEI and SEI.
• Topological indices: chi and kappa series, Hall–Kier alpha, Phi.

Why the profile has three modes

A rule of five read-out is close to useless on a PROTAC, and everyone knows it, yet most calculators still print one and stop. The profile switcher compares a molecule against the space it actually lives in. The oral window follows Lipinski, Veber and Lovering. The fragment window follows the rule of three. The degrader window reflects roughly where reported bifunctional degraders cluster. It describes a crowded region of chemical space, not a strict rule, so treat it as a guide rather than a checklist.

Notes on the numbers

cLogP is a calculated estimate, not a measurement. Wildman–Crippen is an atom contribution model, and on molecules far from its training set (large, zwitterionic or heavily fluorinated ones, and degraders in general), the error can exceed a log unit. Treat it as a ranking tool inside a series, not an absolute value.

TPSA follows the Ertl method as implemented in RDKit and counts nitrogen and oxygen only. You will see small differences against PubChem, which uses a different implementation. Neither is wrong; they're just different definitions, so it's worth knowing which one produced the number in your table.

Counter-ions are stripped by default, so a hydrochloride salt reports the properties of the parent. Turn that off to profile the whole entered structure.

Privacy

The chemistry engine downloads once and then runs inside your browser. Structures are not sent to a server, not logged and not stored. The one exception is name resolution: if you type a name instead of a structure, only that name goes to PubChem. Paste a SMILES and nothing leaves your computer.

Questions

Is this free to use for commercial work?

Yes. Use it for anything, including work you bill for. No account, no limit.

Can I check a whole compound list at once?

Yes. The batch tab takes one SMILES per line and returns a table and a CSV. The work happens locally, so a few thousand structures is a realistic ceiling and depends on your machine.

Why do your values differ from PubChem or ChemDraw?

Different toolkits implement the same descriptor differently. LogP methods disagree by design, and TPSA and rotatable bond definitions vary in what they count. This tool uses RDKit throughout, so your numbers stay internally consistent and reproducible.

Does it calculate pKa or solubility?

No. Both need trained models rather than atom counting, and a wrong pKa is worse than no pKa. If those properties matter for a decision you're about to make, that's worth a conversation, not a guess.

Can I use these descriptors as ML features?

Yes, and the JSON export exists for that. The usual warnings apply: correlated descriptors, distribution shift between your training set and your screening deck, and the pull to trust a model that has never seen a molecule like yours. Curating that data well is most of the work.