A browser-native enzymology textbook that computes.

Philosophy

Modern structural biology is fragmented: beautiful structures on RCSB, mechanisms tucked into M-CSA, kinetics hidden in BRENDA, simulations that demand a cluster and a PhD. This site tries to collapse the distance — one place where you can look at a classical enzyme, step through how it actually catalyses its reaction, read the numbers that define its performance, and compute its slow motions without installing anything.

The scope is intentionally pedagogical and hypothesis-generating. For production-quality molecular dynamics you still want OpenMM, GROMACS, NAMD, or AMBER. For reactive mechanism calculations you want CP2K, ORCA, or Gaussian with QM/MM. What lives here is the layer above — fast, reproducible, shareable, interactive.

What is computed, and how

Structure is rendered with Mol* — the same engine RCSB, PDBe and EMDB use for their viewers — with screen-space ambient occlusion, a subtle outline pass and antialiasing, streamed directly from the RCSB file server. Catalytic residues, mechanism steps and key insights are curated manually from M-CSA and the primary literature.

Dynamics are computed on demand with an Anisotropic Network Model: Cα atoms are linked by Hookean springs within a contact cutoff, the Hessian of the harmonic potential is diagonalised, and the lowest non-trivial eigenmodes describe the protein's collective slow motions. Cross-correlations, betweenness centrality and a steered-response (via the pseudo-inverse) follow from the same eigendecomposition.

The mutation sandbox estimates ΔΔG using a Miyazawa-Jernigan contact potential, with volume-mismatch, charge-flip and class-flip terms scaled by the residue's burial. It is a fast heuristic — not FoldX, not Rosetta — intended to build intuition, not to replace a calibrated calculation.

Nothing about the workbench is AI. All results are deterministic, reproducible, and explainable from the underlying equations. That is the point.

Data sources and licences

• RCSB PDB — structures, public domain.
• UniProt — sequence and functional annotation, CC-BY 4.0.
• M-CSA — curated catalytic mechanisms and residue roles, CC-BY 4.0.
• SABIO-RK — kinetic parameters, CC-BY.
• BRENDA — referenced for additional data (academic licence).
• Primary literature — cited on each enzyme page and in the mechanism references.

All bundled data is compatible with commercial use. If this project is relicensed or integrated into a commercial product, the BRENDA references will be decoupled and the integration will rely on CC-BY sources only.

Limits of the tool

• The workbench ANM assumes a single conformational state. It cannot describe transitions between distinct conformers or unfolding.
• No explicit solvent. Water networks — crucial for enzymes like carbonic anhydrase — are annotated in the mechanism pages but not simulated.
• No reactive chemistry. The ANM is purely harmonic around the crystal structure; it does not make or break bonds.
• Matrix size limits the method to ~1200 residues. Larger systems need coarse-graining or native MD.
• The ΔΔG estimator is a knowledge-based heuristic, not a thermodynamic calculation. Use it to form hypotheses, not to decide.

Next steps

• Expand the catalogue toward 20-50 well-characterised enzymes covering all seven EC classes.
• Pre-computed MD ensembles (mdCATH / BioExcel-CV19) streamed as compressed trajectories.
• On-the-fly pocket detection (Fpocket-style) in the workbench.
• Cryptic-pocket identification from the ANM ensemble.
• QM/MM pre-computed profiles of the catalytic step, as interactive energy diagrams.