PAINS Filter and Substructure Alert Screening Inspector
Inspect chemical structures for Pan-Assay Interference Compounds (PAINS) substructure matches with matched atoms highlighted directly on the 2D depiction. This reference loads mitoxantrone, an approved anthracenedione drug that matches the quinone_A PAINS filter (and anthranil_one_A) in RDKit's PAINS catalogue, to explain why structural alerts require orthogonal biochemical verification rather than blind automated deletion.
Uses the shipped ChEMBL 37 mitoxantrone structure (CHEMBL58), one of 124 approved small-molecule drugs in the ChEMBL 37 library that match at least one published PAINS SMARTS pattern.
Understand how PAINS substructures interfere with biochemical readouts
High-throughput screens often rely on coupled enzymatic reporters, luminescence, fluorescence polarization or singlet-oxygen transfer (AlphaScreen). Certain chemical motifs, such as rhodanines, quinones, catechols, hydroxyphenylhydrazones, curcuminoids and Michael acceptors, repeatedly score as apparent hits across unrelated protein targets because they react non-specifically with cysteine residues, generate hydrogen peroxide via redox cycling in the presence of reducing agents, chelate trace metals or absorb light at the excitation wavelength.
The workspace evaluates the standard RDKit PAINS SMARTS catalog in your browser and shades the exact atoms participating in each matched motif. Seeing which functional group triggered the alert helps a chemist distinguish a peripheral substituent from the core pharmacophore.
Use alerts to prioritise orthogonal controls, not to replace experiments
Journal editors and screening centres rightly require evidence that a reported lead containing a flagged motif engages its target specifically and reversibly. Pair substructure inspection with steep-slope and bell-shape checks in the assay analyser before investing synthetic chemistry resources in an unconfirmed hit.
Equations and variables
Substructure indicator function
IP(G) = 1 if SMARTS pattern P is a subgraph of molecular graph G, else 0
- G
- Standardised two-dimensional molecular graph of the query compound
- P
- Compiled PAINS family A, B or C SMARTS substructure query
- IP(G)
- Binary alert indicator with matched atom indices returned for highlighting
Average molecular weight
MW = Σ ni Mi
- ni
- Count of atoms of element i in the represented structure
- Mi
- Average atomic mass used by the engine (g/mol)
- MW
- Molar mass of the represented structure (g/mol)
Scope and limitations
- PAINS filters were derived from one high-throughput screening library measured across six AlphaScreen assays; they do not detect every promiscuous compound or colloidal aggregator.
- A substructure match identifies a motif associated with historical assay interference (such as redox cycling, singlet-oxygen quenching, fluorescence or non-specific thiol reactivity), not proof that the compound is inactive at its intended target.
- Roughly five to six percent of approved small-molecule drugs (124 of 2,197 in the shipped ChEMBL 37 set) match a PAINS alert, including quinone antineoplastics and azo prodrugs.
Sources and provenance
- RDKit descriptor implementation and versioned reference checksBSD-3-Clause
Method implementation; computed molecular properties are estimates.
- ChEMBL 37 parent structures in the shipped compound libraryCC BY-SA 3.0; share-alike
Existing library capture: 28 September 2026. Preserve attribution when redistributing its rows.
- Wildman and Crippen: atomic contributions to molecular propertiesCited method; no reproduced article data
- Ertl, Rohde and Selzer: fast calculation of molecular polar surface areaCited method; no reproduced article data
- Baell and Holloway: new substructure filters for removal of pan assay interference compounds (PAINS)Cited method; original PAINS family A, B and C definitions
- Aldrich et al.: the ecstasy and agony of assay interference compoundsCited editorial guideline; orthogonal assay verification requirements
Technical questions
Why do 124 approved drugs in the ChEMBL library trigger a PAINS alert?
Many older cytotoxic agents (such as anthracyclines and mitoxantrone), redox-active antiparasitics, dyes and azo-linked anti-inflammatory prodrugs (such as sulfasalazine) contain quinone, catechol or azo motifs whose reactivity or optical absorbance is either part of their pharmacology or tolerated in their specific clinical indication.
Does zero PAINS alerts guarantee that my screening hit is a clean, specific binder?
No. Colloidal aggregation, the single most common source of false positives in biochemical screens, is a physical phase-separation property driven by concentration and lipophilicity rather than a narrow set of reactive functional groups. Many aggregators pass every PAINS filter.
What orthogonal experiments should I run when a hit triggers a PAINS alert?
Test for detergent sensitivity (such as 0.01% Triton X-100) to rule out aggregation, check for redox H2O2 generation or thiol reactivity in the absence of DTT, verify that the compound does not quench or autofluoresce at the readout wavelength, and confirm direct target binding by SPR, ITC or NMR.
Connected workflows
- 4PL IC50 curve fitting
Check whether a flagged screening hit displays an abnormally steep Hill slope.
- Bell-shaped dose-response analysis
Inspect high-concentration assay interference and bell-shaped dose-response artefacts.
- Chemical structure and canonical SMILES
Edit the highlighted substructure motif in the sketcher to see if the alert clears.