Solutions
For pharma and TRPV1 research
Capsaicinoids are TRPV1 agonists with a long clinical history. The platform treats them as a designable chemical family rather than a fixed set of natural products.
TRPV1
primary receptor
Vina
and DiffDock
Boltz-2
structure class
65
production hosts
01
The problem
TRPV1 pharmacology has an unusual shape: a well-characterized receptor, a natural agonist with decades of clinical use, and a chemical family around it that is narrower than it needs to be. The capsaicinoids that exist are the ones peppers happen to make, which is a statement about plant metabolism rather than about chemical space.
Minor capsaicinoids compound the problem. Several are interesting and all are difficult to obtain in quantity, because the plant makes them in trace amounts and separating them from the dominant species is tedious. Biosynthesis that favours a minor species on purpose is a more tractable route than isolation at scale.
02
What the platform does about it
Docking against TRPV1
AutoDock Vina for systematic pose sampling and DiffDock for a learned alternative, both against TRPV1. Two independent methods disagreeing on a candidate is itself informative.
High-confidence models
ESMFold for a fast pass and AlphaFold- or Boltz-2-class prediction where confidence has to carry a decision, rendered in Mol* with per-residue confidence painted onto the model.
Ratio as a design target
The acyl branch sets chain length and saturation, which is what distinguishes the capsaicinoid species. Shifting production toward a minor species is a pathway design problem with a known lever.
Substrate handling
ESM-2 650M scans the pathway enzymes exhaustively, and the shortlist goes to structure. The condensation step at the pun1 locus is where substrate preference is most worth examining.
Host selection
Sixty-five host models with flux analysis behind them, so the question of which organism can actually sustain a route is answered before any of it is built.
Documentation that holds up
Versioned runs with pinned engine versions, ST.26 sequence listings for filings, and checksummed export manifests — built for environments where provenance is scrutinized.
03
A representative project
- Objective
- Biosynthetic route to a minor capsaicinoid
- Receptor context
- TRPV1
- Enzyme scanned
- Capsaicin synthase, pun1 locus
- Scan
- ESM-2 650M, full matrix
- Folded
- ESMFold shortlist, Boltz-2 finalists
- Docking
- AutoDock Vina and DiffDock
- Host modelling
- Flux feasibility across candidates
The work concentrated on the condensation step, because that is where substrate selection between the acyl species is decided. The variant scan produced a shortlist, folding removed the structurally implausible, and docking ordered what remained — with the honest caveat that docking scores are hypotheses to be tested, not measured affinities.
In parallel the flux side answered a question that is easy to defer and expensive to get wrong: which host could actually sustain the route at a useful level. A designed enzyme in a host that cannot supply its precursors is an interesting result and not a usable one.
04
Getting started
The structure gallery and the pathway data are browsable from a free account and are the natural place to begin — the enzymes, their variants, the docking poses already run, each one linked to the computation behind it.
GPU work runs on prepaid credits: $25 for 250 credits, one credit for roughly one Standard job, with Performance and Max classes for structures that need them. Enterprise workspaces add API keys, webhooks, reserved GPU capacity and SSO for groups running this continuously.
Ready to run it
Try this in the Splice Platform.
CPU engines are free to run. GPU engines run on prepaid credits.
Related