A Predicted-Structure Reference for Capsicum Phenylpropanoid and Capsaicinoid Enzymes: 52 Pathway, O-Methyltransferase, Pun1-Variant and Trait-Target Models, and 238 Knockout-Target Structures v1
Year: 2026
Type: preprint
Abstract: Structure prediction has become accurate enough to be useful as a design screen, provided it is read as a screen and not as evidence of function. We report a predicted-structure reference for the enzymes used across the Scoville Splice design packages: 52 Capsicum structures — 14 phenylpropanoid and capsaicinoid pathway enzymes, all 22 members of the caffeoyl O-methyltransferase family, six Pun1 wild-type and variant models, and ten trait targets — together with 238 knockout-target structures across the platform's host species. Three results are design-relevant. The O-methyltransferase set separates the class-I COMT enzymes (~363 aa, acting on free acids) from the class-II CCoAOMT enzymes (242–247 aa, acting on CoA esters), which matters because the methylation step in the capsaicinoid model is a CoA-ester reaction. The Pun1 active-site models place residue 175 in direct contact with the catalytic dyad while S39 and L345 sit in the second shell. An ESM2 substitution scan is reported as a tolerance screen whose ranking is not stable across model size. All structures are ESMFold or AlphaFold2 predictions; none is a statement of activity. § 1 Introduction A predicted structure can tell a designer useful things — whether a modelled fold is plausible, how far a residue of interest sits from a catalytic centre, whether two candidate enzymes share an active-site architecture — without telling them whether the enzyme is active in the host. Treated that way, as a screen, structure prediction sharpens design decisions. Treated as proof of function, it misleads. This reference set is assembled and read in the first sense.
Author: P.I. Ford
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7236479324
