About this dataset
This library holds 2,104 approved small-molecule drugs, one row per active parent compound, taken from ChEMBL 37. 1,713 of them appear as active ingredients in FDA-approved products listed in Drugs@FDA, and the rest were approved by other regulators or are sold in the US without an FDA application. The FDA status column tells you which is which, so you can keep the whole set or narrow it to FDA-approved drugs.
Each row has a canonical isomeric SMILES, ChEMBL and PubChem identifiers, the properties used for drug-likeness filters, the first approval year, an ATC therapeutic class, and the primary mechanism of action. The same drugs are available as a 3D SDF with one minimised conformer each, ready for docking.
What is in the library
ChEMBL marks a molecule as approved (max_phase 4) when a regulator such as the FDA or EMA has authorised a product containing it, which is broader than the US market alone. Salts, hydrates, and solvates were merged into their parent compound, so metformin hydrochloride and metformin are one row with the structure of metformin. Prodrug esters such as abiraterone acetate stay separate from their active forms because ChEMBL treats them as different compounds.
Metal complexes, inorganic salts, gases, contrast-agent mixtures, and anything without a defined structure were left out, because docking engines cannot type those atoms reliably. Antibodies, proteins, and oligonucleotides are not small molecules and are not included. A handful of structurally defined peptides and oligosaccharides under 900 Da, such as carfilzomib, romidepsin, and acarbose, are kept because they are dosed and docked like small molecules.
All descriptors were recalculated with RDKit from the parent structure, so cLogP, TPSA, hydrogen-bond counts, and rotatable bonds are consistent across every row instead of mixing values from different tools.
Using it for virtual screening and drug repurposing
Screening approved drugs against a new target is one of the cheapest ways to find a starting point, because every hit already has known human pharmacology and a supply route. Download the 3D SDF and use it as the ligand set in AutoDock Vina, Smina, or GNINA on ProteinIQ, or send the SMILES to the Boltz small-molecule screen, which predicts binding without a docking box. Individual rows can be sent to a tool directly from the table.
Molecules in the SDF are neutral, in the form ChEMBL records, with explicit hydrogens. They have not been protonated for pH 7.4, so set charge states in your preparation step if your docking engine expects them. Stereocentres that ChEMBL leaves undefined were assigned arbitrarily during embedding, which matters for a small number of racemic drugs.
For a smaller, more drug-like set, filter on the Ro5 violations column (0 keeps drugs that pass all four of Lipinski’s criteria), on molecular weight, or on TPSA. The therapeutic class and mechanism columns help when you want to exclude drugs that already act on your target, or to look for repurposing hits in a particular disease area.
License and how to cite
ChEMBL data is licensed under Creative Commons Attribution-ShareAlike 3.0, so this library is shared under the same license. You can use it for academic and commercial work, including screening campaigns, as long as you credit ChEMBL and ProteinIQ, and any dataset you publish that is derived from it carries the same CC BY-SA terms. Results you obtain by using the data, such as docking scores or hit lists, are not themselves covered by share-alike. Drugs@FDA is a US public domain source and PubChem places no restrictions on its data.
To cite the library, name ProteinIQ’s approved drug library, the version date shown on this page, and ChEMBL 37 (Zdrazil et al., Nucleic Acids Research, 2024, for the ChEMBL database). The ChEMBL ID in every row links back to the source record.
How this dataset was assembled
Drugs come from ChEMBL 37 (released 2026-05-01): every molecule with max_phase 4 was mapped to its ChEMBL parent, so salts, hydrates, and solvates collapse into one row per active parent. Parents typed as small molecules were kept, plus structurally defined peptides and oligosaccharides under 900 Da that ChEMBL types as protein or oligosaccharide (for example carfilzomib and acarbose). Parents without a structure, without carbon, with metals or other elements outside H, B, C, N, O, F, Si, P, S, Cl, Se, Br, and I, or with several disconnected components were excluded, which removes inorganic salts, metal complexes, gases, and mixtures. The FDA status compares each parent's ChEMBL names and synonyms, including its salt forms, with active ingredient names of NDA, ANDA, and BLA products in the Drugs@FDA data files of 2026-10-01; counter-ion words are ignored when no exact name matches. Mechanism is the first disease-relevant mechanism ChEMBL records, therapeutic class is the most common ATC level 2 subgroup across the parent and its salts, and first approval is ChEMBL's earliest approval year. SMILES, formula, molecular weight, Crippen cLogP, TPSA, hydrogen-bond donors and acceptors, rotatable bonds, and Lipinski violations were all recalculated with RDKit 2025.03 from the ChEMBL parent structure. PubChem CIDs were looked up by standard InChIKey and kept only when PubChem returns the same InChIKey for that CID. The 3D SDF has one conformer per drug: hydrogens added, embedded with RDKit ETKDGv3 (fixed seed), and minimised with MMFF94 (UFF where MMFF94 has no parameters). Structures stay in the neutral form ChEMBL records; they are not protonated for pH 7.4, because the rule-based tool we tested assigned wrong charge states to common drugs such as metformin, so set protonation in your docking preparation. 3 drugs whose recorded stereochemistry RDKit could not embed are in the table but not in the SDF.
Each row is one ChEMBL parent compound. Descriptors are recalculated with RDKit from that structure rather than derived from the formula, and every ChEMBL ID and PubChem CID links back to its source record so entries can be checked before reuse.
Sources and further reading
- ChEMBL: Approved drugs (max_phase 4), parent and salt hierarchy, names, mechanisms of action, ATC classes, and first approval years from ChEMBL 37. ChEMBL is licensed CC BY-SA 3.0, so this library and anything derived from it is shared under CC BY-SA 3.0 with attribution to ChEMBL.
- FDA Drugs@FDA: Active ingredients and marketing status of approved US drug products (data files of 2026-10-01), used for the FDA status column. US government work in the public domain.
- PubChem: PubChem CIDs matched by standard InChIKey and checked against PubChem’s own InChIKey. NCBI places no restrictions on PubChem data, though original submitters may retain rights in submitted records.
- RDKit: Canonical SMILES, formulas, descriptors, and 3D conformers (ETKDGv3 and MMFF94).