Key findings
Computed live from the 50-ligand screen and the 1MMQ crystal structure.
Background and context
Why MMP-7 is a target, and why the zinc-binding group matters.
A validated but hard target
The MMP family has at least 28 members, and MMP-1, -2 and -7 are among those experimentally validated as cancer targets. MMP-7 (matrilysin) degrades extracellular-matrix components, and its dysregulated activity is linked to cancer, inflammatory disease and fibrosis.
The hydroxamate legacy
Early broad-spectrum inhibitors used a hydroxamate zinc-binding group to bind the catalytic zinc strongly. They failed in clinical trials; musculoskeletal pain was a common side effect, and lack of selectivity is the usual explanation. Prinomastat, for example, failed a Phase III lung-cancer trial.
The selectivity problem
MMP-1 and MMP-7 have similar S1′ hydrophobic pockets, which makes selective MMP-7 inhibition difficult. Researchers have explored other zinc-binding groups, including a nitro group, and side-chain changes to favour MMP-7 over MMP-1.
Where this project fits
This screen asks how well docked poses reproduce the hydroxamate chelation geometry seen in the crystal structure, and how that compares with other chemotypes. Only MMP-7 was docked, so selectivity across the family is outside what these results can show.
The catalytic zinc machinery
In PDB 1MMQ the catalytic Zn²⁺ is held by three histidines (His218, His222, His228 of the HEXXHXXGXXH motif) plus both hydroxamate oxygens at 2.16 and 2.20 Å, with Glu219 hydrogen-bonding the N–OH. Any credible pose must satisfy this geometry, not just score well.
Why it still matters
MMP-7 remains a biomarker and therapeutic candidate in cancer and kidney fibrosis, with recent high-resolution structures (e.g. 8JUD/8JUF/8JUG, 1.3–1.5 Å) renewing interest in selective, non-hydroxamate chemotypes. That is why this screen judges affinity and zinc-chelation fidelity as separate evidence.
Sources: RSC Advances 2015, nitro-based MMP-7 inhibitors · Dove Press review of MMP inhibitors · Patsnap Synapse overview of MMP-7 inhibitors (a general overview, less rigorous than the other two).
Screen at a glance
All 50 ligands ranked by predicted affinity. Click a tile to open that ligand in the Ligand explorer.
Workflow
Click a stage to see what was done and which tools were used.
Tools per analysis layer
Filter by stage.
| Tool | Stage | Role |
|---|
Zn²⁺-mediated inhibition
Bidentate chelation of the catalytic zinc by the hydroxamate group, with three coordinating histidines.
Pharmacophore interaction network
Force-directed schematic of the binding-site relationships. Drag nodes, click one for its role.
Click a node.
Large filled = Zn²⁺ · filled = protein residue · outlined = ligand · faint = pocket. Solid line = coordination; dashed = H-bond; dotted = hydrophobic.
3D active-site explorer
Drag to rotate. Real atomic coordinates from PDB 1MMQ (1.90 Å): the catalytic Zn²⁺ (centre), the co-crystallised hydroxamate inhibitor (outlined atoms) and the 16 residues within 4.5 Å of it. Dashed lines are metal coordination.
Active-site anatomy (PDB 1MMQ)
Measured directly from the crystal structure coordinates. Residue numbers follow the PDB entry.
Zn²⁺ coordination shell
| Atom | Distance to Zn²⁺ |
|---|---|
| His218 NE2 | 2.16 Å |
| Inhibitor O2 | 2.16 Å |
| Inhibitor O1 | 2.20 Å |
| His228 NE2 | 2.24 Å |
| His222 NE2 | 2.31 Å |
Five atoms within 2.4 Å: three histidines and both hydroxamate oxygens, i.e. bidentate chelation. The second zinc lies about 10 Å from the inhibitor.
Polar contacts (≤ 3.4 Å)
| Protein atom | Inhibitor atom | Å |
|---|---|---|
| Glu219 OE1 | O1 | 2.57 |
| Tyr240 N | O4 | 2.76 |
| Leu181 N | O3 | 2.91 |
| Ala182 O | N1 | 2.91 |
| Asn179 O | N3 | 2.92 |
| Pro238 O | N2 | 3.01 |
| Glu219 OE2 | N1 | 3.03 |
Seven closest of 13 contacts; several involve backbone atoms, so they are not side-chain specific.
Residues within 4.5 Å of the inhibitor
The zinc-binding motif is His218-Glu219-Leu220-Gly221-His222, with His228 as the third zinc ligand.
Screening results
Fifty known MMP-7 actives from ChEMBL (target CHEMBL4073) docked into the catalytic site of PDB 1MMQ with AutoDock Vina.
Ranked predicted affinity (kcal/mol)
Solid bars: a ligand oxygen lies within 2.6 Å of the catalytic Zn²⁺. Faint bars: no such contact.
Affinity against closest ligand O–Zn²⁺ distance
Predicted affinity against measured potency
| # | Ligand | Affinity | Measured pChEMBL | Min O–Zn (Å) | O ≤ 2.6 Å | Zn contact | Bidentate-like | ZBG class |
|---|
Vina’s default scoring has no explicit metal term, so affinity and Zn²⁺ geometry are reported as separate evidence. Predicted scores are not measured potencies.
Zinc-binding group analysis
Ligands grouped by chemotype. Classes are assigned by SMILES pattern matching on the terminal group, not a full substructure search.
| Class | Ligands | Zn contact | Bidentate-like | Mean affinity | Best affinity |
|---|
Predicted affinity by zinc-binding group
More visualisations
Further views of the screen and of the 1MMQ crystal contacts. Solid colour means a Zn²⁺ contact; faint means none.
Affinity distribution
Closest O–Zn²⁺ distance distribution
Size vs affinity, and ligand efficiency
Ligand metric heatmap
Ligands ranked by affinity. Darker means a better score, a shorter O–Zn distance, or the flag is set.
1MMQ crystal contact map
Distance (Å) from each pocket residue to each atom of the co-crystallised inhibitor. Darker is closer; numbers shown up to 4.0 Å.
Additional views
Group comparisons, ranking behaviour, multi-metric profiles and the crystal ligand’s distance to the zinc.
Affinity by class
O–Zn²⁺ distance by class
Boxes show quartiles and the median; whiskers show the range; dots are individual ligands. The phosphinate class has a single ligand.
Zn²⁺ contact rate down the ranking
Share of ligands with a Zn²⁺ contact among the top k ranked by affinity; dashed line is the overall rate.
Class → binding outcome
Multi-metric profiles (parallel coordinates)
One line per ligand for the top ligands chosen by the ligand slider above. Thicker lines are bidentate-like.
Crystal inhibitor atoms by distance to the catalytic Zn²⁺ (1MMQ)
Further visualisations
Correlations between metrics, radial and density views of the screen, and the crystal pocket as a contact wheel, a sequence map and a distance-shell histogram.
Metric correlation matrix
Pearson r across the 50 ligands. Darker means stronger. Affinity is negative, so a negative r with size means larger ligands score stronger.
Affinity density by class
Radial ranking
Top ligands from the ligand slider; longer bars are stronger predicted binding. Solid: Zn contact; faint: none. Rings mark −6, −7, −8 and −9 kcal/mol from the centre outwards.
Inhibitor contact wheel (1MMQ)
Residues from the residue slider around the co-crystallised inhibitor. Spoke width is the number of atom pairs within 4.0 Å; solid nodes make a polar contact within 3.4 Å.
Pocket residues along the sequence
Taller stems are closer to the inhibitor; rings mark the three zinc-coordinating histidines. The axis is not to scale: // marks a gap in the sequence.
Atoms around the catalytic Zn²⁺ (1MMQ)
Heavy atoms per 0.5 Å shell: protein N/O (solid), other protein atoms (medium), inhibitor atoms (faint).
Structure overview (PDB 1MMQ)
Backbone geometry, flexibility, secondary structure and pocket composition computed directly from the crystal coordinates and header records.
Ramachandran plot
Each point is one residue (termini excluded). αR, β and αL mark the usual regions approximately; no allowed-region contours are drawn.
Flexibility and secondary structure along the sequence
Cα B-factor per residue. Ringed markers are the 16 pocket residues; rounded bars below the axis are helices and arrows are strands, taken from the PDB header.
Pocket composition by side-chain class
Hit finder: score and chelation fidelity
Set a score cutoff and how close the closest oxygen must sit to the crystal Zn²⁺–O distance (2.18 Å, the mean of 2.16 and 2.20 Å). Ligands passing both criteria are highlighted.
Score against chelation geometry: interactive hit selection
| # | Ligand | Affinity | Min O–Zn (Å) | Class |
|---|
Crystal Zn²⁺ coordination geometry (1MMQ)
| Donor–Zn–donor angle | His218 | His222 | His228 | Inhib. O1 | Inhib. O2 |
|---|---|---|---|---|---|
| His218 | – | 99.5° | 104.2° | 100.2° | 94.1° |
| His222 | 99.5° | – | 97.5° | 88.2° | 163.4° |
| His228 | 104.2° | 97.5° | – | 153.7° | 88.4° |
| Inhib. O1 | 100.2° | 88.2° | 153.7° | – | 79.9° |
| Inhib. O2 | 94.1° | 163.4° | 88.4° | 79.9° | – |
Five donors (three histidine NE2 atoms and both hydroxamate oxygens). The geometry index τ5 is 0.16, so the site is a distorted square pyramid (0 is an ideal square pyramid, 1 an ideal trigonal bipyramid). The hydroxamate bite angle O1–Zn–O2 is 79.9°; the two largest angles, His222–O2 (163.4°) and His228–O1 (153.7°), set the base of the pyramid.
Docked poses can only be compared with this geometry once their coordinates are available; the screen here records O–Zn distances.
Design tools
Three interactive tools for exploring the screen: build your own ranking, find the Pareto front between any two metrics, and compare two ligands head to head.
Custom ranking designer
Weight what matters to you. Each metric is scaled 0 to 1 across the 50 ligands, then combined; bars show each metric’s contribution to the top 10.
Chelation closeness is 1 when the closest oxygen sits at the crystal distance (2.18 Å) and falls to 0 at 1.5 Å away.
Pareto front explorer
Pick any two metrics. Better is always up and to the right. Ligands on the front are not beaten on both metrics by any other ligand.
| Ligand | Class |
|---|
Head-to-head comparison
Orbit map and contact links
Two alternative views: the screen arranged around the catalytic zinc, and the crystal inhibitor–residue contacts drawn as links.
Zinc orbit map
Distance from the centre is each ligand’s closest O–Zn²⁺ distance (rings at 2.18 Å crystal, 2.6 Å cutoff, then 3–6 Å; the radial scale is a square root so the crowded 2–3 Å region is spread out). Ligands are grouped by class (hydroxamates first, then acid series, then phosphinate) and ordered by affinity rank around the circle. Dot size is heavy-atom count; solid dots have a Zn contact; outlined dots are bidentate-like.
Inhibitor–residue contact links (1MMQ)
Each link joins an inhibitor atom to a pocket residue within 4.0 Å; thicker, darker links are closer. Residues follow the residue slider above.
Ligand explorer
Pick any docked ligand to see its metrics, where it sits in the screen, and its structure as a SMILES string.
Reference structures
Experimental MMP-7 structures from the Protein Data Bank (checked against RCSB, PDBe and NCBI records).
MMP7 (also MPSL1, PUMP1) · P09237
The smallest member of the matrix metalloproteinase family
Complexes typically carry two Zn²⁺ and two Ca²⁺ ions (e.g. 8JUF)
The zinc-coordinating group dominates the relative potency of inhibitors
| PDB ID | Entry | Resolution | Source |
|---|---|---|---|
| 1MMQ | Matrilysin complexed with a hydroxamate inhibitor (single 170-residue chain) | 1.90 Å | Browner et al., Biochemistry 1995 |
| 1MMP | Matrilysin complexed with a carboxylate inhibitor | 2.30 Å | Browner et al., Biochemistry 1995 |
| 8JUD | Human MMP-7 with a peptide inhibitor | 1.50 Å | Deposited 2023-06-26 |
| 8JUF | Human MMP-7 with a peptide inhibitor | 1.39 Å | Abe-Sato et al., J Med Chem 2023 (MMP-7 inhibitors for kidney fibrosis) |
| 8JUG | Human MMP-7 with a peptide inhibitor (175 residues, UniProt 95–267) | 1.30 Å | PDBe |
| 8K4Z | Human MMP-7 with an inhibitor | 1.70 Å | Deposited 2023-07-20 |
All entries solved by X-ray diffraction. The 1995 study also reported a sulfodiimine complex (2.3 Å maximum resolution) and compared how each zinc-binding group engages the catalytic zinc.
Docking Lab
Click Load example to see it work on the top 10 ligands from the screen above, or paste your own results as name, score (kcal/mol). Optionally add H-bonds, hydrophobic, salt bridges, Zn contact (0/1) to get a fingerprint heat-table. Runs entirely in your browser.
How to run your own screen
The same pipeline used for the results above, step by step.
PDB 1MMQ (matrilysin with a hydroxamate inhibitor). The script downloads it, keeps chain A with Zn²⁺ and Ca²⁺, and centres the grid box on the co-crystallised inhibitor.
Make a CSV with name,smiles. This screen used the 50 most potent known MMP-7 actives from ChEMBL target CHEMBL4073 (best pChEMBL per molecule).
python dock_mmp7.py --pdb 1MMQ --ligands ligands_mmp7.csv --exhaustiveness 8 --confs 3
Needs Open Babel, RDKit, Meeko and Vina. Each ligand is minimized, docked, and its best pose is measured against the catalytic zinc. No setup? Run the Colab notebook or the Streamlit lab instead.
Paste lab_paste.txt into the Docking Lab above for a ranked chart. Read results.csv for O–Zn distances and contact flags.
Open poses/<ligand>_complex.pdb in PyMOL. For residue-level contacts run plip -f poses/<ligand>_complex.pdb -t or use LigPlot.
Judge affinity and Zn²⁺ geometry as separate evidence, using the guide below.
How to read the outputs
| Column | Meaning | How to read it |
|---|---|---|
| affinity_kcal_mol | Best Vina pose score | More negative means stronger predicted binding. Use it to rank, not as an absolute energy. |
| min_O_Zn_A | Closest ligand oxygen to the catalytic Zn²⁺ | In the 1MMQ crystal structure the hydroxamate oxygens sit at 2.16 and 2.20 Å. Values near 2.2–2.4 Å resemble real chelation. |
| n_O_within_cutoff | Oxygens within 2.6 Å of Zn²⁺ | Counts how many oxygens reach the metal. |
| zn_contact | 1 if at least one oxygen is within 2.6 Å | A pose that touches the zinc at all. |
| bidentate_like | 1 if two or more oxygens are within 2.6 Å | The two-point chelation seen for hydroxamates. |
Worked example: CHEMBL41497
Best score of the 50 (−9.25 kcal/mol), closest oxygen 2.3 Å from Zn²⁺, and two oxygens within the cutoff, so it is bidentate-like. That matches the crystal chelation distances of 2.16 and 2.20 Å, so the pose reproduces the expected binding mode.
Counter-example: CHEMBL5572317
Second-best score (−9.05) but its closest oxygen is 2.65 Å from the zinc, just outside the cutoff, so it is not counted as a contact. A strong score does not guarantee chelation, which is why both are reported.
Interaction analysis
Residue-level recognition with PyMOL, LigPlot and PLIP.
Methods in depth
What each step contributes and its main limitation.
| Method | Contribution | Limitation |
|---|---|---|
| Energy minimization | Removes strain and clashes so ligands start from sensible geometries. | Depends on the force field; minima are not necessarily bioactive conformers. |
| Multi-conformer generation | Samples alternative geometries before screening. | Coverage grows with count, as does cost. |
| AutoDock Vina | Fast pose sampling and affinity ranking. | Scores are approximate; the default scoring has no explicit metal-coordination term, so Zn²⁺ geometry must be checked manually. |
| PyMOL / LigPlot | 3D and 2D views of contacts. | Visual judgement is partly subjective. |
| PLIP | Automated, geometry-based interaction typing. | Fixed distance and angle cut-offs; static snapshots only. |
Glossary
Terms used on this page.