Skip to content

Methods and checksMethods

How every result is reached and what it was checked against, app by app. The same as each app's Methods dialog.

Methods and checks pages

Drybench Molecule: methods and checks

How Drybench Molecule reaches its results: every calculator, analysis or view with its methods, the references behind them, how far each has been checked, and the app's limits.

Open Drybench Molecule · What the levels mean · Every app's methods and checks

Scorecard

One molecule with tabs for its overview, properties, similar drugs, chemical space and 3D shape.

Its result goes through 19 methods. The lowest level among them: Tested, not reference-checked.

Below the top level: Lipinski's rule of 5 with Crippen logP: fails when more than one of MW ≤ 500, logP ≤ 5, historical Lipinski N/O hydrogens ≤ 5 and N/O atoms ≤ 10 is broken (Tested, not reference-checked); Veber's rules: 10 or fewer rotatable bonds and a polar surface area of 140 Ų or less (Tested, not reference-checked).

Methods

Average molecular weight from natural isotope abundances (RDKit amw)

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Monoisotopic mass (RDKit exactmw)

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Wildman-Crippen atom-contribution estimate of octanol/water log P (RDKit CrippenClogP); not BioByte's CLOGP

  • Level: Checked against a reference
  • Reference: Wildman SA, Crippen GM. J Chem Inf Comput Sci 1999;39:868-873. doi:10.1021/ci990307l
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Topological polar surface area from N and O contributions (RDKit tpsa)

  • Level: Checked against a reference
  • Reference: Ertl P, Rohde B, Selzer P. J Med Chem 2000;43:3714-3717. doi:10.1021/jm000942e
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Hydrogen-bond donor sites (RDKit NumHBD); historical Lipinski counts are separate

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Hydrogen-bond acceptor sites (RDKit NumHBA); historical Lipinski N/O counts are separate

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Lipinski's historical donor count: hydrogens on N and O (RDKit lipinskiHBD, NHOHCount); the rule of 5 is judged on it

  • Level: Checked against a reference
  • Reference: Lipinski CA, Lombardo F, Dominy BW, Feeney PJ. Adv Drug Deliv Rev 1997;23:3-25. doi:10.1016/S0169-409X(96)00423-1
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Lipinski's historical acceptor count: N and O atoms (RDKit lipinskiHBA, NOCount); the rule of 5 is judged on it

  • Level: Checked against a reference
  • Reference: Lipinski CA, Lombardo F, Dominy BW, Feeney PJ. Adv Drug Deliv Rev 1997;23:3-25. doi:10.1016/S0169-409X(96)00423-1
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Rotatable bonds, RDKit strict definition (NumRotatableBonds)

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Ring count (symmetrised SSSR; bridged cages count extra rings)

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Rings whose atoms are all aromatic

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Fraction of carbons that are sp3

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Atoms other than hydrogen

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Net formal charge as drawn

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Quantitative estimate of drug-likeness, weighted (QED.py ported from RDKit)

  • Level: Checked against a reference
  • Reference: Bickerton GR, Paolini GV, Besnard J, Muresan S, Hopkins AL. Nat Chem 2012;4:90-98. doi:10.1038/nchem.1243
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

PAINS substructure filters A, B and C (480 SMARTS, RDKit FilterCatalog definitions)

  • Level: Checked against a reference
  • Reference: Baell JB, Holloway GA. J Med Chem 2010;53:2719-2740. doi:10.1021/jm901137j
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Morgan fingerprints (RDKit; radius 2, 2,048 bits, connectivity invariants, bond types used, no chirality, no counts) on the described fragment, policy morgan-r2-2048-described-v1, compared by Tanimoto similarity

  • Level: Checked against a reference
  • Reference: Rogers D, Hahn M. J Chem Inf Model 2010;50:742-754. doi:10.1021/ci100050t
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Lipinski's rule of 5 with Crippen logP: fails when more than one of MW ≤ 500, logP ≤ 5, historical Lipinski N/O hydrogens ≤ 5 and N/O atoms ≤ 10 is broken

Veber's rules: 10 or fewer rotatable bonds and a polar surface area of 140 Ų or less

  • Level: Tested, not reference-checked
  • Reference: Veber DF, Johnson SR, Cheng HY, Smith BR, Ward KW, Kopple KD. J Med Chem 2002;45:2615-2623. doi:10.1021/jm020017n

How it is made and tested

  • Map: A principal component analysis of nine standardised descriptors over the 2,197 approved small-molecule drugs of ChEMBL 37.
  • Engine: RDKit.js 2026.03.6 (WebAssembly) in a Web Worker; OpenChemLib in a second worker for the 3D conformer.

Compare

The molecule among the library's approved drugs, with the drugs it most resembles side by side.

Its result goes through 19 methods. The lowest level among them: Tested, not reference-checked.

Below the top level: Lipinski's rule of 5 with Crippen logP: fails when more than one of MW ≤ 500, logP ≤ 5, historical Lipinski N/O hydrogens ≤ 5 and N/O atoms ≤ 10 is broken (Tested, not reference-checked); Veber's rules: 10 or fewer rotatable bonds and a polar surface area of 140 Ų or less (Tested, not reference-checked).

Methods

Average molecular weight from natural isotope abundances (RDKit amw)

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Monoisotopic mass (RDKit exactmw)

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Wildman-Crippen atom-contribution estimate of octanol/water log P (RDKit CrippenClogP); not BioByte's CLOGP

  • Level: Checked against a reference
  • Reference: Wildman SA, Crippen GM. J Chem Inf Comput Sci 1999;39:868-873. doi:10.1021/ci990307l
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Topological polar surface area from N and O contributions (RDKit tpsa)

  • Level: Checked against a reference
  • Reference: Ertl P, Rohde B, Selzer P. J Med Chem 2000;43:3714-3717. doi:10.1021/jm000942e
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Hydrogen-bond donor sites (RDKit NumHBD); historical Lipinski counts are separate

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Hydrogen-bond acceptor sites (RDKit NumHBA); historical Lipinski N/O counts are separate

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Lipinski's historical donor count: hydrogens on N and O (RDKit lipinskiHBD, NHOHCount); the rule of 5 is judged on it

  • Level: Checked against a reference
  • Reference: Lipinski CA, Lombardo F, Dominy BW, Feeney PJ. Adv Drug Deliv Rev 1997;23:3-25. doi:10.1016/S0169-409X(96)00423-1
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Lipinski's historical acceptor count: N and O atoms (RDKit lipinskiHBA, NOCount); the rule of 5 is judged on it

  • Level: Checked against a reference
  • Reference: Lipinski CA, Lombardo F, Dominy BW, Feeney PJ. Adv Drug Deliv Rev 1997;23:3-25. doi:10.1016/S0169-409X(96)00423-1
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Rotatable bonds, RDKit strict definition (NumRotatableBonds)

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Ring count (symmetrised SSSR; bridged cages count extra rings)

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Rings whose atoms are all aromatic

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Fraction of carbons that are sp3

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Atoms other than hydrogen

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Net formal charge as drawn

  • Level: Checked against a reference
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Quantitative estimate of drug-likeness, weighted (QED.py ported from RDKit)

  • Level: Checked against a reference
  • Reference: Bickerton GR, Paolini GV, Besnard J, Muresan S, Hopkins AL. Nat Chem 2012;4:90-98. doi:10.1038/nchem.1243
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

PAINS substructure filters A, B and C (480 SMARTS, RDKit FilterCatalog definitions)

  • Level: Checked against a reference
  • Reference: Baell JB, Holloway GA. J Med Chem 2010;53:2719-2740. doi:10.1021/jm901137j
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Morgan fingerprints (RDKit; radius 2, 2,048 bits, connectivity invariants, bond types used, no chirality, no counts) on the described fragment, policy morgan-r2-2048-described-v1, compared by Tanimoto similarity

  • Level: Checked against a reference
  • Reference: Rogers D, Hahn M. J Chem Inf Model 2010;50:742-754. doi:10.1021/ci100050t
  • How it was checked: Checked against Python RDKit 2026.03.6 on every approved drug in the library (pipeline run of 10 October 2026): every value matches; the largest QED difference is 2.8e-16; the described-fragment bits of the 75 multi-fragment rows against the fingerprint-policy fixture of 10 October 2026.

Lipinski's rule of 5 with Crippen logP: fails when more than one of MW ≤ 500, logP ≤ 5, historical Lipinski N/O hydrogens ≤ 5 and N/O atoms ≤ 10 is broken

Veber's rules: 10 or fewer rotatable bonds and a polar surface area of 140 Ų or less

  • Level: Tested, not reference-checked
  • Reference: Veber DF, Johnson SR, Cheng HY, Smith BR, Ward KW, Kopple KD. J Med Chem 2002;45:2615-2623. doi:10.1021/jm020017n

How it is made and tested

  • Map: A principal component analysis of nine standardised descriptors over the 2,197 approved small-molecule drugs of ChEMBL 37.

What else is checked

  • Expected outputs from Python RDKit 2026.03.6 (descriptors, QED, PAINS matches and Morgan bits) for reference molecules. The browser's results are compared with it.
  • The fixture and RDKit.js are the same RDKit release; descriptors, QED, PAINS and Morgan bits match Python RDKit; invalid and oversized input is refused; salts, metal complexes and unspecified stereocentres are handled; a tautomer of a library drug has the same standard InChIKey.
  • Tanimoto on packed bits equals set Tanimoto in Python RDKit, nearest-neighbour ranking, the PCA and the stored map projection, Lipinski and Veber, deep links, PubChem property names, the library file's licence and release, route flags and formula ordering.

Data sources

Limits

Not medical advice. Nothing here is advice about any medicine, treatment or condition, and nothing here is an assessment of any person's variant, sample or molecule. Decisions made with it are yours.

Not a forecast. Stages, scores, rates, rankings and similarities shown here describe records as they stand, or historical averages; they are not a forecast for any molecule, trial, drug, company or target.

  • For research and education only. It is not a prediction of whether any molecule will work, be absorbed or be safe, and it is not medical advice.
  • Standard InChI normalises selected tautomer classes, not all tautomerism: 2-pyridone/2-hydroxypyridine can share a key, while ordinary ketone/enol drawings need not. Fragment selection is a separate app-specific parent-selection policy and does not make every salt or drawing identical.
  • Generic HBD/HBA use RDKit NumHBD/NumHBA. Historical Lipinski N/O hydrogen and atom counts are used separately for Rule-of-Five logic. Strict rotatable-bond and TPSA definitions are RDKit implementations of the cited heuristics, not exact reproductions of every original descriptor system.
  • Rules of thumb and QED describe typical oral drugs. Many approved drugs break them (natural products, antibiotics, drugs taken up by transporters, injectables), so a fail is a prompt to look closer, not a verdict. They are not given for metal complexes or inorganic salts, which they were never fitted to.
  • Crippen logP and TPSA are fast calculated estimates (the Wildman-Crippen and Ertl methods), not measurements; Crippen logP is not the BioByte CLOGP that Lipinski used. The rotatable-bond count uses RDKit's strict definition, which can differ from the software used in the original papers.
  • A salt or mixture is described by its largest organic fragment, with its charge as drawn: the tool does not neutralise it. If a smaller fragment holds a metal other than a simple counter-ion such as sodium, potassium, magnesium or calcium, the whole input is kept instead.
  • The 3D shape is one conformer from a single seeded run, minimised in vacuum. Real molecules move between many shapes, and a bound drug may look different. Stereocentres not defined in the input (racemic drugs such as ibuprofen and thalidomide, for example) are given an arbitrary configuration in 3D. Molecules with over 100 heavy atoms (in the largest fragment) are not built, large or ring-rich ones (cyclosporin, for example) can take a minute or more, and metal complexes are not modelled.
  • The library holds approved parent molecules that ChEMBL types as small molecules and that have a structure. Drugs without a structure (many metal complexes, such as cisplatin, and mixtures) are missing. So are drugs ChEMBL types as proteins, oligosaccharides or unknown, which include peptides and some small molecules such as theophylline, ciclosporin, carfilzomib and acarbose.
  • Route groups come from ChEMBL's oral, parenteral and topical flags, which mark routes a drug is known to be given by. They are records, not predictions, and the tool does not say whether any molecule would be orally active. A missing flag means ChEMBL has none, not that the route is never used. ChEMBL counts inhaled and eye products as topical (salmeterol, latanoprost). A flag set only on a salt or other form counts for the drug; this adds one, bortezomib. Of the 2,197 drugs, 1,268 are oral, 371 parenteral but not oral, 221 topical only and 337 other or unknown.
  • 75 library drugs are stored by ChEMBL as salts (sodium benzoate, for example) and are compared as stored, so a salt can score below 1.00 against its own parent.
  • Approval years are ChEMBL's first_approval field. For many older drugs this is the US approval year, which can be decades after the drug was first used elsewhere: metformin shows 1995, but it was first reported as a diabetes treatment in France in 1957 (Bailey CJ, Diabetologia 2017;60:1566-1576). Some drugs have no year, including many approved only outside the US (gliclazide and domperidone) and a few approved in the US (bortezomib, approved in 2003); some drugs approved only outside the US do have one (agomelatine, 2009). 268 of the 2,197 drugs are flagged by ChEMBL as withdrawn for toxicity in at least one country; they stay in the library and on the map.
  • A saved molecule keeps the source it came from and a date. For a curated GSRS structure that date is when the record was curated, not when you looked it up; for a ChEMBL lookup it is when your browser received the answer, which this site may have kept from an earlier request, usually less than a day before but possibly longer.
  • The map keeps 67% of the variance in the nine descriptors (42% on the first axis, 25% on the second), so points close on the map can still differ in ways the map does not show. It compares properties, not structures or targets.