STRUCTURE → HYPOTHESIS → EXPERIMENT

Fewer guesses.
Better evidence
for your experiment.

Choose mutations for your first enzyme stabilization experiments. Compare structures, examine risks and prepare a construct design.

Explore a protein ↗

AlphaFold DB · UniProt · FireProtDB
Reproducible calculations. Testable hypotheses.

01 / ENZYME STABILITYValidate predictions in the laboratory

WORKSPACE

From a question to the first experiment series

Checking server…

INPUT DATAUniProt

Enzyme stabilization

Choose your protein and a homolog to compare. Confirm the donor’s property independently.

Your protein / reference

Homolog to compare

Expression systemCandidates

Have a CDS? Add your DNA

The full coding sequence for the selected UniProt entry: A/C/G/T or a single FASTA record. A terminal stop codon is allowed.

Open FASTA

Without a CDS, primer designs refer to a new synthetic gene. For an existing plasmid, supply your DNA sequence.

EXPLORE AN EXAMPLE

Saved data open immediately and are available without the server.

WORKED EXAMPLE

Subtilisin → thermitase

Style

Background

HighlightOverlay homolog

HOW THE RESULT IS PRODUCED

Every recommendation
needs evidence.

We distinguish structure geometry, stability predictions and experimental evidence. Missing data are never treated as confirmation.

Compare and check

Sequence alignment, structure superposition and a pLDDT confidence filter. Functional sites are excluded from the candidate list.

Rank hypotheses

A linear Foldiff model performs the main selection. When configured, ThermoMPNN adds a separate prediction. Disagreements are shown explicitly.

Prepare testing

Spatial contacts, primer designs, data exports and calculation parameters. Confirm mutation and combination effects experimentally.

Model metrics and limits

Precision@20 — fraction of stabilizing substitutions among the top twenty predictions in saved splits. These metrics do not guarantee results for a new enzyme. The model does not predict temperature or activity gains or pH optimum shifts. Spatial separation does not rule out epistasis. AlphaFold structures are predictions.

FireProtDB lookup finds reported measurements; it is not independent validation of a model trained on that database. Different constructs, numbering and assay conditions require manual review.

FOR LABORATORIES

Start with your
research question.

A pilot needs a protein, experimental goal, expression system and any measured variants. The deliverable is a reasoned plan for the first experiment series.

PILOT DELIVERABLES

Question → calculation → testing

Candidates with rationales and limits. Review of disagreements with reported measurements. A construct design using agreed DNA.

Public beta without registration

Current server limits appear above the form

For research use

Discuss a pilot: @oligoplan_bot. Confidential data require a private laboratory account.

Fewer guesses.Better evidencefor your experiment.