Data
Pathway map
A flux solution is a long list of numbers until you draw it. Escher maps render a COBRApy solution as a metabolic network with the load visible on every edge — which is usually when the bottleneck becomes obvious.
COBRApy
solver
Escher
renderer
65
host models
~90%
flux on one branch
01
What the visualization is
Flux balance analysis answers a budget question: given what the organism takes in and what it must keep doing to stay alive, how much carbon can reach the end of a particular branch? COBRApy computes it. Escher draws it, with reaction edges weighted by the flux they carry, so a solution reads as a picture of where the material actually goes.
The maps are specific to the host. Each of the 65 organisms carries its own metabolic model, with the capsaicinoid branch native in the peppers and grafted on everywhere else. What the map shows is that host’s metabolism under that design, not a generic diagram.
02
How to read it
Edge weight is flux
Thick edges carry load, thin edges carry little, and edges at zero are inactive under this solution. The shape of the active subnetwork is usually more informative than any single number in it.
Bottlenecks are narrow
A branch that constricts at one reaction is the branch worth engineering. In the capsaicinoid pathway that constriction is almost always on the vanillylamine side, which carries around ninety percent of flux control.
Competition is visible
Precursor pools feed more than one destination. The map shows what the capsaicinoid branch is competing against, which explains targets that look achievable in isolation and are not.
Comparison by overlay
Two solutions can be rendered against each other so the difference is the thing you see. This is the fastest way to understand what a design change actually did to the budget.
Everything is clickable
A reaction opens its bounds, its gene associations and the enzymes behind it, with links through to variant scans and structures for those enzymes.
Assumptions stay attached
Media composition, objective function and constraints are shown with the map. A flux map without its assumptions is decoration, and the platform does not present one that way.
03
A worked look
- Host
- Tomato (Solanum lycopersicum)
- Model
- Genome-scale, fruit-tissue constraints
- Objective
- Capsaicinoid branch flux
- Target
- 2,500 SHU, Tier 2
- Verdict
- Feasible
- Bottleneck
- Vanillylamine branch, PAMT step
- Overlay
- Against SS-0421 v4 solution
Tomato is the instructive case because it has no capsaicinoid pathway at all. The map shows the phenylpropanoid intermediates it does make, the designed branch hanging off them, and the narrow point where the amine supply runs out. Everything downstream of that constriction has capacity to spare — which is exactly why tuning a downstream enzyme would have been wasted effort.
Overlaying the earlier solution made the revision legible in one glance: the branch widened, the competing sink shrank slightly, and the overall budget held. That is a successful change, and it is far easier to see than to summarize numerically.
04
What you can learn from it
Whether a heat target is affordable in a given host. Which single reaction to reinforce, and which ones not to bother with. What a design change cost elsewhere in metabolism. Whether two hosts that look similar on paper have genuinely similar budgets, or only similar diagrams.
Flux analysis and map rendering are CPU engines and run free, which makes iteration cheap. Most designs run dozens of solutions before anything is committed, and that is the intended use.
Ready to run it
Try this in the Splice Platform.
CPU engines are free to run. GPU engines run on prepaid credits.
Related