How we design a Scoville Splice cultivar

01

Transcriptomics

Every design starts in the placental tissue of a superhot pepper fruit. We RNA-sequence this tissue across developmental timepoints and across Capsicum chinense accessions ranging from habanero to Carolina Reaper.

From the expression matrix we recover the full capsaicinoid-pathway module: Pun1 / AT3, pAMT, KAS, BCAT, and the vanillin-arm oxidases. Co-expression networks pinpoint which transcription factors we need to carry alongside the structural genes.

02

Flux-balance analysis

With the gene set in hand we build a stoichiometric model of the capsaicinoid pathway and run FBA. The model reveals that in nearly every host, the vanillylamine branch — not capsaicin synthase itself — is the rate-limiting step.

FBA also sets our ceiling SHU per host. A tomato fruit, with a 20-day ripening window and a modest acyl-CoA pool, caps near 150,000 SHU. Rice endosperm, with a far smaller precursor pool, caps near 20,000. These numbers determine which Tier a product can occupy.

03

AI protein engineering

We score every single-amino-acid substitution in Pun1 and pAMT with ESM2, Meta FAIR's 650-million-parameter protein language model. ESM2 ranks mutations by predicted stability and activity without needing any wet-lab iteration.

Top-ranked variants are stacked into enzyme sets per Tier. Pepper Tier variants use near-wildtype Pun1; Tier 3 and above use re-engineered Pun1 that out-performs the native sequence on branched acyl-CoA substrates.

04

CRISPR constructs

For each product we computationally design guide RNAs covering both the knock-out of host competing branches and the knock-in of the capsaicinoid cassette. See CRISPR and construct for background.

Delivery is through a T-DNA binary vector for most hosts, with ribonucleoprotein editing available for lines where transgene-free status matters.

05

Capsaicinoid blend chemistry

SHU alone does not define a chili's character. The ratio among capsaicin, dihydrocapsaicin, and nordihydrocapsaicin determines whether the burn is front-of-mouth, back-of-throat, slow-build, or short-and-sharp.

For each product we compute a target five-capsaicinoid ratio and tune promoter strengths across the pathway to hit it. The result is a reproducible burn signature — the difference between a Pyroclast and a Kindling at the same nominal SHU.

06

DNA assembly

Every design is lowered to a synthesis-ready DNA sequence: codon-optimized coding sequences, host-matched promoters, terminators, selection markers, and homology arms sized to the chosen insertion locus.

The final construct ships as a signed bundle: FASTA, JSON manifest, FBA predictions, CRISPR guides, datasheet, and short product paper. 31,911 base pairs of ready-to-order DNA per product; one million base pairs across the full catalog.