Written and maintained by CASRAI Editorial Board
Published
On this page
The PubChem Sketcher is the structure-drawing editor built into PubChem, the open chemistry database run by the National Center for Biotechnology Information (NCBI). You draw a molecule (or paste a SMILES, SMARTS, InChI or InChIKey string that the Sketcher converts into a drawing), click Search for This Structure, and PubChem returns compounds that are identical to, similar to, or contain your drawing as a substructure.
This guide covers what the Sketcher is used for in research, how to open it, a step-by-step search, how the four structure search types differ, how to read the hit list, and what to check when a search returns nothing useful. The settings quoted below come from the NCBI PubChem structure search help, the Current Protocols chapter on PubChem structure search and an NLM training exercise, all listed under Sources. For the wider database, see the PubChem guide.
What the PubChem Sketcher is used for in research
The Sketcher answers questions that a name search cannot. A name or identifier search needs you to already know what the compound is called. A structure search needs only the molecule you have in mind, or part of it. Typical research uses, all supported by the PubChem documentation, are:
- Confirming identity. Check whether the exact compound you drew is already a PubChem record and find its compound identifier (CID).
- Finding analogues. Retrieve compounds whose structure is close to a query, using a similarity threshold you control.
- Finding a scaffold. Retrieve every compound that contains a core fragment, such as a ring system, with a substructure search. The Current Protocols chapter gives the example of CID 15207492 being a substructure of CID 3961, which is losartan.
- Going from structure to data. The chapter’s Basic Protocol 4 draws a query in the Sketcher, runs a substructure search and then downloads the bioactivity data of the hits.
How to open the PubChem Sketcher
The Sketcher is part of the PubChem website at pubchem.ncbi.nlm.nih.gov; there is nothing to install. The NLM training exercise and the Current Protocols chapter describe three routes in:
- Draw Structure on the homepage. The button opens the Sketcher. The top text box beside it also accepts a SMILES, SMARTS, InChI or InChIKey string directly.
- From a compound you already found. Search for the compound, open it under Best Match, and click Similar Structures Search and Edit Structure. The Sketcher opens with that compound drawn, ready to modify.
- Structure search page. The structure search help lists the Identity/Similarity and Substructure/Superstructure tabs, and the Sketcher is opened from them with the Launch button.
Step-by-step: draw a structure and search PubChem
The steps below follow the NLM exercise and the Current Protocols chapter. Menu labels are quoted as those documents give them.
- On the PubChem homepage choose Draw Structure, or open the Sketcher from a compound page as described above.
- Build the query. Use the tool buttons and the modified periodic table to place atoms and bonds. To change hydrogens, the NLM exercise points to the Hydrogens pull-down and its Add special option. Alternatively, paste a line notation into the text box of the Edit Structure window and press Enter, and the Sketcher draws it. The exercise uses salicylic acid written as
O(C1=CC=CC=C1C(O[H])=O)[H], and its acetylated versionO(C1=CC=CC=C1C(O[H])=O)C(C)=O, as examples. - Click Search for This Structure. The chapter notes that the query can be drawn manually or converted from a line notation such as a SMILES or InChI string.
- Choose the search type: identity, similarity (2-D or 3-D), substructure or superstructure. The chapter states that identity is shown first by default.
- If you need a different threshold or matching option, open Settings and change it, then run the search again (details in the next section).
- Read the hit list (see below), and use Search All if the default cap truncated it.
Identity, similarity, substructure and 3-D: which search to run
PubChem offers four structure search types from the Sketcher. The data table below sets out each type’s scoring basis, default setting and adjustable range, taken from the NCBI help and the Current Protocols chapter.
Identity search
An identity search finds records that are the same compound as the query. By default a hit must share connectivity, isotopes and stereochemistry with the query, so a successful search usually returns only the compound itself; the NLM exercise likewise reports that a successful identity search typically returns one compound record. The help describes seven identity options: Any Tautomer; Same Connectivity; Same Stereoisomer; Same Isotopic Labels; Same Stereochemistry and Isotopes (the recommended exact match); Same Isotopes and Nonconflicting Stereoisomers; and Nonconflicting Stereoisomers. Choose a looser option, such as Same Connectivity, when you want stereoisomers or isotopologues of the same skeleton as well.
Similarity search
A 2-D similarity search ranks compounds by the Tanimoto coefficient on PubChem’s binary substructure fingerprint, which has 881 bits. The help adds that this similarity ignores stereochemistry and isotopes, and that a 100% match is an exact match ignoring stereo and isotopes. The default threshold is 90% (0.9 in the chapter); Settings lets you lower it, with predefined thresholds from 99% down to 60%. A higher threshold returns fewer, closer hits.
Substructure and superstructure search
A substructure search returns compounds that contain your drawing; a superstructure search returns compounds that are contained within it. The substructure options listed in the help are: Match stereochemistry (Ignore, Exact, Relative or Nonconflicting); Match isotope (off by default); Match charge (off by default); Match tautomers; Ringsystems not embedded; Single/double bonds match aromatic bonds (on by default); Chain bonds match ring bonds (on by default); and Strip hydrogen.
3-D similarity search
The 3D Conformer tab searches precomputed conformers, up to three per compound. The default cutoffs are a shape similarity of 80% and a feature (color) similarity of 50%, written in the chapter as shape-Tanimoto at least 0.80 and color-Tanimoto at least 0.50.
Output and how to read it
The results page lists matching compound records. What a hit means depends on the search type you ran, so read the list accordingly:
- Identity: expect one record, or a very short list if you loosened the options. Each hit is the same compound under the matching rule you chose.
- Similarity: each hit meets or exceeds the Tanimoto threshold, so a 90% hit is a close analogue and not necessarily the same scaffold. Lowering the threshold to 60% widens the list sharply.
- Substructure: each hit contains your drawn fragment; the hits can be much larger molecules than the query.
By default each type of structure search stops when a maximum of 1000 hit compounds is found. The Search All option extends the search, with a cap of 1,000,000. If your hit list stops at exactly 1000, treat it as truncated and either run Search All or make the query more specific. From a hit you can open the compound record or, as in the chapter’s Basic Protocol 4, download bioactivity data for the hits. The Sketcher also has an export function that can create images of your structure, as the NLM exercise notes. Compounds returned here can be cross-checked against other resources such as the DrugBank guide, the ChemSpider guide or the ChEMBL dictionary entry.
Troubleshooting
An identity or similarity search returns nothing or the wrong thing
The help states that identity and similarity searches need completely defined structures, and recommends explicit hydrogens for them. Wildcards or SMARTS patterns in an identity or similarity query may give unintended results. Use a fully specified structure for those two search types, and keep query patterns for substructure searches.
The exact compound is not found but a related one exists
The default identity rule requires matching connectivity, isotopes and stereochemistry. If the compound is recorded without stereochemistry, or as another tautomer, choose Same Connectivity or Any Tautomer instead of the exact-match option.
Too many substructure hits
Tighten the query by adding atoms, or use the stereochemistry, isotope and charge matching options. The chapter also notes that substructure matching is limited to 1000 hits unless you choose Search All.
A SMILES or InChI will not draw as expected
Paste the string into the Edit Structure text box and press Enter so the Sketcher converts it, and check the drawing before searching. The search help lists SMILES, SMARTS (with * as a wildcard), InChI, a CID, a molecular formula and an uploaded SDF file (V2000 or V3000) as the accepted inputs on the structure search tabs.
Free alternatives and complements
| Option | What the source says | Source |
|---|---|---|
| PubChem search box (no drawing) | Accepts a SMILES, SMARTS, InChI or InChIKey string typed or pasted into the top text box on the homepage. | NLM exercise |
| ChEBI | An open-access database and ontology of chemical entities that can be queried with text and structure searches, for example by name, molecular formula, InChI or SMILES. | EMBL-EBI ChEBI |
Frequently asked questions
What is the PubChem Sketcher?
It is PubChem’s structure editor. You draw a molecule or convert a SMILES or InChI string into a drawing, then search PubChem for identical, similar, substructure or superstructure matches.
What does the PubChem Sketcher search by?
The search types are identity, 2-D similarity, 3-D similarity, substructure and superstructure. Identity is shown first by default.
What is the default similarity threshold?
The default is 90% Tanimoto similarity (0.9). Settings lets you lower it to as little as 60%.
Why does my search stop at 1000 compounds?
Each structure search type stops at 1000 hits by default. Search All extends the search up to 1,000,000.
Can I paste a SMILES string instead of drawing?
Yes. SMILES, SMARTS, InChI and InChIKey strings can be pasted into the text box and the Sketcher converts them to a drawing.
Sources
- NCBI PubChem, structure search help: https://pubchem.ncbi.nlm.nih.gov/search/help_search.html (read 2026-11-14)
- Current Protocols, PubChem structure search chapter: https://pmc.ncbi.nlm.nih.gov/articles/PMC8363119 (read 2026-11-14)
- NLM workshop, Introduction to PubChem, exercise 2: https://www.nlm.nih.gov/ncbi/workshops/2023-06_intro-to-pubchem/exercise-2.html (read 2026-11-14)
- EMBL-EBI ChEBI: https://www.ebi.ac.uk/chebi/ (read 2026-11-14)
Data: PubChem structure search types: scoring basis, defaults and adjustable settings
Compares the four PubChem structure search types available from the Sketcher by scoring basis, default setting, adjustable range and default hit cap, compiled from the NCBI search help, the Current Protocols chapter and an NLM training exercise.
| Search type | Scoring basis | Default setting | Adjustable | Default hit cap |
|---|---|---|---|---|
| Identity | Exact-match rule over connectivity, stereo and isotopes | Same connectivity, isotopes and stereochemistry | Seven identity options (for example Any Tautomer, Same Connectivity) | 1000 hits; Search All up to 1,000,000 |
| 2-D similarity | Tanimoto on 881-bit PubChem substructure fingerprint (ignores stereo and isotopes) | 90% (0.9) | Threshold 99% down to 60% via Settings | 1000 hits; Search All up to 1,000,000 |
| Substructure / superstructure | Query contained in (or containing) the hit | Isotope and charge matching off; aromatic and ring bond matching on | Stereo (Ignore, Exact, Relative, Nonconflicting), isotope, charge, tautomers, strip hydrogen | 1000 hits; Search All up to 1,000,000 |
| 3-D similarity | Shape and feature (color) Tanimoto on up to 3 precomputed conformers | Shape at least 0.80, color at least 0.50 | Defaults stated; adjustable range not stated in the sources | 1000 hits; Search All up to 1,000,000 |
Sources
- PubChem structure search help (accessed 2026-11-14)
- Current Protocols: PubChem structure search (accessed 2026-11-14)
- NLM workshop exercise 2 (accessed 2026-11-14)
Compiled by CASRAI from the sources above. Free to reuse under CC BY 4.0 with attribution.







