Chemistry and compounds · What you can do here
Calculate molecular properties from SMILES
Calculate molecular descriptors and assess drug-likeness rules from SMILES. Receive standardised structures, physicochemical properties, and rule violations without local installations.
What this involves
Calculating molecular properties from a SMILES string requires rigorous chemical standardisation before running any descriptor calculations. Recalling SMILES strings or running descriptors directly on uncurated inputs regularly introduces silent errors, particularly when counterions, salt adducts, or ambiguous stereocentres distort calculated molecular weights and lipophilicity values.
This workflow evaluates common medicinal chemistry metrics including molecular weight, Wildman–Crippen cLogP, topological polar surface area (TPSA), hydrogen bond donors and acceptors, rotatable bonds, ring counts, and fraction of sp3 carbons (Fsp3). In addition to individual physicochemical properties, the agent benchmarks inputs against established heuristics such as Lipinski's Rule of Five, Veber's parameters, and lead-likeness criteria.
Calculations are handled deterministically using RDKit on the platform's backend environment. When working with chemical names rather than confirmed strings, the agent verifies identity against curated databases such as ChEMBL and PubChem, strips extraneous fragments, neutralises charges, and flags assay interference alerts using the PAINS and Brenk catalogues.
How it goes
- 1Structure verification and retrievalIf starting from a compound name or synonym, the agent queries ChEMBL (falling back to PubChem) via search_compounds to resolve the canonical SMILES, formula, and full 27-character InChIKey.
- 2Fragment stripping and charge standardisationThe agent parses the molecule using RDKit, isolating the parent species with rdMolStandardize.LargestFragmentChooser to remove counterions and applying rdMolStandardize.Uncharger before computing properties.
- 3Descriptor calculationUsing molecule_properties or custom RDKit routines via run_python, the agent computes molecular weight, Wildman–Crippen cLogP, TPSA, rotatable bonds, Fsp3, aromatic ring counts, and stereocentre numbers.
- 4Drug-likeness and alert profilingThe computed metrics are evaluated against Lipinski Ro5, Veber, and lead-likeness rules, alongside PAINS (PAINS_A/B/C) and Brenk substructure triage via FilterCatalogParams.
Start it here
Ask in plain language. The agent works on a Linux machine of its own — it installs and runs what the job needs, keeps every file, and reads the results back with the numbers explained. Nothing to set up, and a long run continues after you close the page.
What to watch for
- Never write a SMILES string from memory; a recalled string with an incorrect bond order or stereocentre silently invalidates all downstream metrics.
- Salts add between 36 and 230 Da; calculating descriptors without stripping counterions like HCl or TFA corrupts molecular weight, cLogP, and rule compliance.
- A SMILES string lacking directional bonds or @/@@ tags has undefined stereochemistry, which is chemically distinct from a racemic mixture or a planar system.
- Never treat Wildman–Crippen MolLogP as experimental logD7.4; logD7.4 for basic compounds frequently sits two to three units below calculated neutral logP.
- PAINS alerts are triage flags for screening lists rather than definitive verdicts; roughly 5% of approved drugs trigger PAINS substructure matches.
- Never use canonical tautomers when reporting a structure, as the canonical tautomer generated by RDKit is a computational convention rather than a determined solution state.
What it runs
Models
Databases it can query
Questions
Do I need to install anything to calculate molecular properties from smiles?
No. The research agent does this work on a Linux machine of its own — it installs what the job needs, keeps the files, and reads the results back to you. Long runs continue if you close the page.
Why does the calculated cLogP differ from values published in the literature?
RDKit computes cLogP using the Wildman–Crippen atom-contribution model, which is fitted to neutral structures and has an RMSE around one log unit, with wider divergence on zwitterions, strong bases, and internal hydrogen bonds. Furthermore, published papers typically report experimental logD7.4, which can be two to three log units lower than logP for basic compounds.
Can I provide a compound name instead of a SMILES string?
Yes. The agent uses search_compounds to query ChEMBL and PubChem for verified names or synonyms, returning canonical SMILES, formulas, clinical phases, and InChIKeys.
How are structural tautomers and salts handled before property calculation?
Counterions are removed using RDKit's LargestFragmentChooser and structures are uncharged to prevent salt mass from inflating molecular weight and lipophilicity. Tautomers are canonicalised only when performing structural comparisons, never when reporting the primary compound.
Does failing Lipinski's Rule of Five mean a molecule should be discarded?
No. Lipinski's Rule of Five is an empirical observation regarding orally dosed compounds, allows one violation, and does not apply to natural products, macrocycles, peptides, PROTACs, or actively transported drugs. Violations are reported to highlight profile liabilities without automatically rejecting the molecule.
Written from the method the research agent follows for this work and the documentation of the tools listed above. Statements citing a figure that does not appear there are removed before publication. Corollary is a non-commercial research demonstration; results are not for clinical use.