Structures and potency
Bound is not necessarily potent
Fourteen protein sites across eleven families. Every co-crystal ligand solved sitting in the pocket, put into one frame. Their measured activity spans six orders of magnitude.
The ladder
Every dot is one ligand solved in that protein's site, placed by its measured potency. Color runs from the weakest value in the whole set to the strongest, one scale for every row. Click a row to open that site in the viewer.
The number at the right is the fold difference in potency: the concentration the weakest ligand needs divided by what the strongest needs for the same effect. In HSP90 that is 759 nanomolar against 0.48 nanomolar, about 1,585 fold. In PDE4D it is 11 micromolar against 20 picomolar, about 562,000 fold. All of these compounds crystallized in the site.
One site, many molecules
Every co-complex of a protein superposed on the CA atoms of its own pocket. Sticks are colored on that site's potency scale, red at its weakest measured ligand and green at its strongest, with unmeasured ligands in gray. Purple spheres are ordered waters. Drag to rotate.
Consensus threshold
Show
Potency, this site
Key
A acceptor, D donor, N negative, P positive, R aromatic ring, H hydrophobic. A point with more than one feature carries all its letters.
A ligand id shown as a link is a compound the Kinase Knowledgebase has measured. The link opens the KKB search with that structure loaded in the editor.
| PDB | Ligand | Resolution | CA RMSD | Measured p | Source |
|---|
Every site
Potency is on a negative log molar scale: 9.00 is one nanomolar, 6.00 is one micromolar. Span is the gap in log units between the weakest and strongest measured ligand in a site, and fold difference is 10 raised to the span. A span of 3.00 is a thousandfold, 6.00 a millionfold. Every ligand here was solved at 1.8 angstroms or better in the same pocket.
| Protein | Family | Ligands superposed | With a value | Weakest | Strongest | Span | Fold difference |
|---|
How this was built
Structures come from the RCSB PDB at 1.8 angstrom resolution or better and R-free 0.23 or better, restricted to drug-like ligands of 150 to 800 daltons passing Lipinski with no violations.
Superposition. For each protein, the highest-resolution structure is the reference and is left exactly as deposited. Every other structure is fitted onto it by a Kabsch rotation and translation of its protein, using the CA atoms within 20 angstroms of the reference ligand, paired by residue number, with at least 12 pairs and a CA RMSD of 2.5 angstroms or better. That same rotation and translation is then applied to the ligand of that structure. The ligand is never moved relative to its own protein; it arrives in the reference frame because its protein was fitted there. Where an entry holds more than one copy of the ligand, every copy is carried across and the one landing closest to the reference pocket is kept, within 6 angstroms of the reference ligand's center. CA RMSD in the table is the residual of that fit, and the reference row shows reference in its place.
Pharmacophore typing is pfpall's. A receptor atom reaches a ligand at its feature's own distance: 3.5 angstroms for a hydrogen bond, 4.0 for an ionic pair, 4.5 for hydrophobic contact, 5.5 for an aromatic ring centroid.
Potency is the strongest exact measurement for that compound against that protein, from ChEMBL 37 for every family and from the Kinase Knowledgebase Q2 2026 release for kinases: pIC50, pKi, pKd or pEC50. Compounds are matched on the InChIKey connectivity block and the UniProt accession, so salt, protonation and stereochemical variants resolve to one compound. Families are ChEMBL's protein classification.
A co-complex shows that a molecule occupied a site under crystallization conditions. Potency is a separate measurement, and within one pocket the two differ by up to six orders of magnitude.
What each kinase ligand is actually doing → Contact maps with typed, directional interactions measured from the deposited coordinates, for every Kinase Knowledgebase compound with a matching PDB structure.Talk to us about the Kinase Knowledgebase
The kinase rows here sit on top of KKB structure activity data, the same data behind the Kinase Foundation Model. We are happy to walk through the method, the coverage and what it is useful for.