Enzyme–substrate workbench
Catalytic geometry & mutagenesis
Start with your enzyme and target substrate, or upload existing structures to compare their geometry.
Start with a mutation walkthrough
Try a complete mutation case
Start with a real stored WT–mutant pair. Inspect the hypothesis and mutation site, compare geometry, then export a single-mutant sequence and analysis report.
Loading replaces the current draft and comparisons. No account, credits or GPU run is needed. Source-model geometry does not establish improved activity.
Save or reopen a design draft
Save your sequence, goal, context, chemistry inputs and mutation settings in a local file. Structures, confidence files and measurement recipes are separate exports. No draft is sent to our server.
Where your inputs go across the workflowDescribe your design goal
Enter your sequence, target substrate and desired outcome. Add any useful experimental context below.
Add context, constraints or prior results · optional
Exporting saves your inputs; prediction runs separately in Kaggle.
Use a reaction example
Choose an example to fill in candidate molecules. Add your own enzyme sequence.
Start the chorismate mutase benchmark
Loads the 127-residue 2CHT sequence, substrate, trimer hypothesis and design context, replacing the design inputs above. Chemical identity and assembly still need your review before export.
No prediction is started. This reference binds a transition-state analogue; it is not a chorismate-bound structure.
Prepare cofolding
Review your molecules, export a plan, then run the prediction in Kaggle.
Substrate: Not provided
When the requirement is unknown, a cofactor code above is a candidate to investigate. Example presets supply candidate molecules; they do not establish the chemistry of your uploaded sequence.
Check chemical states and molecule counts
Specify the reviewed charges and molecule counts for the whole enzyme assembly. Resolve unknown cofactor or metal requirements before exporting.
Set modeling pH and cofactor requirement above. Counts are explicit and independent of protein copies. A CCD code may encode a different charge from the relevant state at your pH; use a verified charged SMILES when needed.
How chemistry checks work
Two Mg²⁺ per active site in a dimer with two occupied sites means four ions. Use enzyme-specific evidence; salt concentration and charge balance do not determine catalytic ion counts.
This records your reviewed hypothesis; an agent has not independently verified the evidence. Before prediction, the Kaggle notebook uses RDKit to check the actual molecule charges, undefined stereochemistry and metal ion identities against this plan. It does not infer pKa values or physiological protonation from pH alone.
Run cofolding on Kaggle
Download the notebook and import it into Kaggle. Enable GPU and Internet, paste your reviewed plan into the input cell, then run it. Start with one pose.
Free GPU availability and quotas depend on Kaggle. The notebook runs when you start it; exporting inputs does not submit a job. It uses the external MSA service for protein sequences by default. Download and extract the results ZIP, then upload the predicted .cif files in Compare structures. Keep your notebook and inputs private when needed.
Run on other compute
The reviewed plan contains a Boltz input document. Apply the same chemistry checks when running it on other compute. The MSA-server option sends protein sequences to the external MSA service. Upload the predicted .cif files in Compare structures. Treat forced reactive-distance constraints as modeling assumptions, not independent evidence that the reaction geometry was predicted.
Boltz input, compute and confidence documentationMeasure catalytic geometry
Upload your WT and mutant complexes as PDB or mmCIF files. Choose WT as the reference, then add the distances or angles you want to compare.
Build a focused mutation library
Choose pocket residues using the complex and your reaction mechanism. All positions below are 1-based positions in the sequence you pasted above. Map structure numbering first. Each exported variant changes one residue.