Workflow
Design a cultivar from scratch
Six stages take a blank project to a construct-ready specification — organism in, annotated DNA and a predicted heat rating out. This is the path every one of the 1,000 catalog entries travelled.
6
stages
65
organisms
1,000
reference constructs
100 – 16M
SHU range
01
The goal
A cultivar design is a hypothesis written in DNA. You are asserting that a particular set of coding sequences, promoters and regulatory edits, placed in a particular host, will produce a particular concentration of capsaicinoids in a particular tissue. The designer exists to make that hypothesis explicit, testable and portable — something a wet lab can act on without reconstructing your reasoning from scratch.
The starting point is deliberately small: a host organism from the 65 in the catalog, a heat target somewhere on the ladder from 100 SHU to 16,000,000 SHU, and a burn profile describing how that heat should arrive and fade. Everything downstream is derived, versioned and inspectable.
02
Stage by stage
Frame the target
Choose the host, the tissue, the SHU target and the capsaicinoid ratio you want. The designer immediately shows the nearest catalog neighbours so you can see whether you are extrapolating or interpolating.
Model the flux
A COBRApy metabolic model of the host is loaded and the capsaicinoid branch is grafted on. Flux balance analysis tells you whether the carbon and nitrogen budgets support the target, and Escher renders the result as a readable map.
Design the enzymes
Meta ESM-2 650M scores substitutions across the pathway enzymes, with ESMFold and Boltz-2 structures for the variants worth a closer look. You are picking candidates, not single answers.
Plan the edits
CHOPCHOP- and CRISPOR-style guide scoring proposes edit sites; Cas-OFFinder sweeps the host genome for off-target matches so a guide with a crowded neighbourhood never reaches the construct.
Assemble the construct
DNA Chisel handles codon adaptation for the host, Primer3 proposes verification primers, and pydna plus SBOL turn the parts list into an assembly record rather than a loose pile of sequences.
Validate and export
The compliance pass checks the design against the project's rules, then the export bundles GenBank, FASTA, SBOL and vendor-ready order sheets with a manifest describing exactly what produced them.
03
Worked example
- Project
- SS-0421 — Tier 2 tomato
- Host organism
- Tomato (Solanum lycopersicum)
- Target tissue
- Fruit pericarp
- Heat target
- 2,500 SHU
- Burn profile
- Front-loaded, short tail
- Pathway branch
- Vanillylamine via PAMT
- Series
- Garden Ladder
Tomato has no capsaicinoid pathway of its own, so the model starts from the phenylpropanoid intermediates it does make and asks what the fruit can spare. Flux balance analysis puts a ceiling on the branch well before any enzyme work begins — which is the useful part. A target that the host cannot fund is better discovered in an afternoon of modelling than in a season of greenhouse space.
The vanillylamine branch governs roughly ninety percent of flux into the pathway, so that is where the design effort concentrates: PAMT activity on the amine side, and capsaicin synthase (the pun1 locus product) on the acyl side. The designer surfaces that asymmetry rather than treating every enzyme as equally worth tuning.
04
What you get
A finished project is a complete, dated record: the model and its assumptions, the variant shortlist with scores, the guide set with its off-target evidence, the assembled construct in GenBank and SBOL, verification primers, and an export manifest naming every engine version involved. Hand it to a collaborator and they can see what you decided and why.
CPU engines — flux modelling, codon work, primers, annotation, assembly — run free. GPU engines draw on prepaid credits: $25 buys 250 credits, and one credit is roughly one Standard job, with Performance and Max classes available when a structure deserves the extra compute. Most designs spend their credits on structures and variant scoring and nothing else.
Ready to run it
Try this in the Splice Platform.
CPU engines are free to run. GPU engines run on prepaid credits.
Related