ChEMBL FAQ

Q: ChEMBL vs PubChem -- when should I use each?

ChEMBL and PubChem are complementary but serve different purposes: Rule of thumb: Use ChEMBL when you need measured bioactivity data with target assignments. Use PubChem when you need compound identifiers, vendor availability, or computed molecular properties.

Q: What is the ChEMBL data license?

ChEMBL data is released under the Creative Commons Attribution-ShareAlike 3.0 Unported (CC BY-SA 3.0) license. This means: You can freely use, share, and adapt ChEMBL data for any purpose, including commercial use You must give appropriate credit to ChEMBL/EMBL-EBI If you redistribute modified data, you must use the same license There are no restrictions on using ChEMBL data in drug discovery pipelines, publications, or commercial products, as long as attribution is provided.

Q: What do compound confidence scores mean?

ChEMBL assigns a confidence score (1-10) to each compound-target association, reflecting how directly the target was identified in the original assay: Recommendation: For SAR analysis and medicinal chemistry decisions, filter for confidence ≥ 7. For broader target landscape surveys, confidence ≥ 4 is acceptable.

Q: How often is ChEMBL updated?

ChEMBL releases new versions approximately 2-3 times per year. Each release adds newly curated compounds and bioactivity data from recent publications. The platform uses the ChEMBL REST API, which always serves the latest release. Release notes are published at chembl.gitbook.io and include statistics on new compounds, targets, and assays added in each version.

Frequently asked questions about ChEMBL bioactivity data, licensing, and comparisons with other databases.

These answers cover when to reach for ChEMBL versus PubChem, the bioactivity data types (IC50, Ki, EC50, Kd, pChEMBL), the CC BY-SA 3.0 data license, what confidence scores and target classifications mean, structure-based searching, and why the same compound-target pair can show multiple activity values.

ChEMBL Frequently Asked Questions

ChEMBL vs PubChem -- when should I use each?

ChEMBL and PubChem are complementary but serve different purposes:

Aspect	ChEMBL	PubChem
Data source	Curated from peer-reviewed medicinal chemistry literature	Deposited by labs, vendors, and automated extraction
Strength	High-quality bioactivity data (IC50, Ki, EC50) with confidence scores	Broad compound coverage with computed properties
Compound count	Millions, curated from literature	Tens of millions, aggregated from many sources
Best for	SAR analysis, target-based drug design, hit validation	Compound lookup, vendor sourcing, property calculation
Curation	Manual by EMBL-EBI curators	Mostly automated

Rule of thumb: Use ChEMBL when you need measured bioactivity data with target assignments. Use PubChem when you need compound identifiers, vendor availability, or computed molecular properties.

What are the different bioactivity data types?

ChEMBL reports several types of activity measurements. The most common:

Type	Full Name	What It Measures	Typical Use
IC50	Half-maximal inhibitory concentration	Concentration to inhibit 50% of activity	Enzyme and cell-based inhibition assays
EC50	Half-maximal effective concentration	Concentration for 50% of maximum effect	Agonist and functional assays
Ki	Inhibition constant	Equilibrium dissociation constant for an inhibitor	Competitive binding assays (enzyme kinetics)
Kd	Dissociation constant	Equilibrium binding affinity	Biophysical binding assays (SPR, ITC)

Key distinction: IC50 depends on assay conditions (substrate concentration, incubation time), while Ki is a thermodynamic constant independent of assay setup. Ki values are more comparable across studies, but IC50 is far more commonly reported.

ChEMBL also standardizes values as pChEMBL (-log10 of the molar value), which allows comparison across activity types on a single scale. Higher pChEMBL = more potent.

What is the ChEMBL data license?

ChEMBL data is released under the Creative Commons Attribution-ShareAlike 3.0 Unported (CC BY-SA 3.0) license. This means:

You can freely use, share, and adapt ChEMBL data for any purpose, including commercial use
You must give appropriate credit to ChEMBL/EMBL-EBI
If you redistribute modified data, you must use the same license

There are no restrictions on using ChEMBL data in drug discovery pipelines, publications, or commercial products, as long as attribution is provided.

What do compound confidence scores mean?

ChEMBL assigns a confidence score (1-10) to each compound-target association, reflecting how directly the target was identified in the original assay:

Score	Description	Reliability
9	Direct single protein target	Highest -- biochemical assay on purified protein
8	Homologous single protein target	High -- inferred from a closely related species
7	Direct protein complex target	High -- multi-subunit target directly tested
6	Homologous protein complex	Moderate -- complex target inferred from homolog
5	Direct selectivity measurement	Moderate -- panel assay with selectivity data
4	Cell-based, mechanism-based	Lower -- target inferred from mechanism
1-3	Phenotypic or uncertain	Low -- target assignment uncertain

Recommendation: For SAR analysis and medicinal chemistry decisions, filter for confidence ≥ 7. For broader target landscape surveys, confidence ≥ 4 is acceptable.

How are targets classified in ChEMBL?

ChEMBL organizes targets into a hierarchy based on protein classification:

Single protein: A specific protein (e.g., EGFR, CHEMBL203). Most common and most reliable for SAR.
Protein complex: A multi-subunit target (e.g., GABA-A receptor). Activity data reflects the complex, not individual subunits.
Protein family: A group of related proteins (e.g., "Matrix metalloproteinases"). Less specific but useful for selectivity analysis.
Cell-line: Activity measured in a cell line without a specific molecular target assigned.
Organism: Whole-organism assay (e.g., antimicrobial MIC values).

Each target also has a target type (e.g., kinase, GPCR, ion channel) from the ChEMBL target classification system, which maps to standard protein families.

Can I search ChEMBL by structure instead of name?

Yes, by similarity: provide a SMILES string and a Tanimoto threshold (e.g., 70%) to find structurally similar compounds. See Similarity Search. There is no separate ChEMBL substructure-search tool on the platform — for a specific motif, run a similarity search at a lower threshold and filter by hand, or use PubChem's substructure search instead.

For exact compound lookup by structure, convert your molecule to an InChIKey and search by that identifier for unambiguous matching.

How often is ChEMBL updated?

ChEMBL releases new versions approximately 2-3 times per year. Each release adds newly curated compounds and bioactivity data from recent publications. The platform uses the ChEMBL REST API, which always serves the latest release.

Release notes are published at chembl.gitbook.io and include statistics on new compounds, targets, and assays added in each version.

Why do I see different IC50 values for the same compound-target pair?

Multiple IC50 values for the same compound-target pair are expected and normal. Common reasons:

Different assay formats: Biochemical vs. cell-based assays yield different IC50 values
Different laboratories: Experimental conditions (temperature, buffer, substrate concentration) vary
Different publications: Each measurement comes from an independent study

To handle this, ChEMBL does not merge or average values. Instead, use the confidence score and assay description to select the most relevant measurement for your analysis. For systematic comparisons, consider using the median or geometric mean of available values.