Expression atlas

A gene being present in a genome and a gene being expressed in the tissue you care about are different facts, and designs fail on the gap between them more often than on anything exotic. The atlas is the platform’s answer to that: transcriptomic data assembled across all 65 catalog organisms and arranged so expression can be read by tissue, by developmental stage, and by gene.

Coverage is uneven and the atlas says so rather than smoothing it over. The Capsicum species are densely sampled — fruit, leaf, root, flower, across ripening — because that is where the capsaicinoid pathway lives natively and where the catalog began. Tomato and tobacco are well covered as the workhorse cross-species hosts. The long tail, from soybean to raspberry to vanilla, is thinner, and the interface shades sparse regions instead of letting them look authoritative.

Question

Will a seed-specific promoter carry the pathway?

Construct

SS-0613 — Miso Burn, Bean Counter series

Host

Soybean (Glycine max)

Target

2,500 SHU, Tier 4

Promoter

Seed-specific

Genes filtered

PAMT, capsaicin synthase, 9 others

Axes

Tissue × developmental stage

Comparison host

Capsicum chinense, fruit

Cost

Free to browse

Soybean is an awkward host for a pathway that evolved in fruit, and the atlas is where that awkwardness becomes concrete. Filtering to the eleven genes in the construct and switching the axis to developmental stage shows the seed-specific profile turning on late and sharply — which is the behaviour the design wants, and not at all what a constitutive promoter would have given it.

Putting habanero fruit beside it on the same scale is the useful part. The native profile is broader and earlier; the soybean profile is narrower and later; the overlap is the window the design actually has to work in. That window, not the SHU figure, is what determined the promoter choice on this construct — and the same comparison against vanilla pod tissue is why SS-0900 went a different way entirely.

Whether a host expresses the precursor machinery a target depends on, or only carries the genes. Whether a promoter fires in the right tissue at the right stage. Whether two hosts that look comparable on a phylogeny behave comparably in the tissue that matters, which they frequently do not. Where the data is thin enough that a design is an extrapolation rather than an inference.

It is also the fastest way to kill a bad idea. A pathway can be perfectly feasible in a flux solution and still be pointless if the host only expresses the relevant machinery in roots. Checking the atlas takes a minute; discovering it in a greenhouse takes a season.

Cultivar designers reach for it before committing to a promoter. Researchers use it to decide which of the 65 hosts is worth a serious attempt. Teams running comparative work across a series — Bramble, Orchid, Bean Counter — keep it open beside the genome browser, because the two views answer halves of the same question.

Browsing the atlas is free and carries no credit cost. Credits only come into it when a question turns structural and you move on to variant scoring or folding, and by then the atlas has usually settled which host and which tissue you are arguing about.