Ultimately, we could have a full gene-expression-as-a-dynamical-system model which we could perturb and use to study the dynamical features of genetic networks.
It looks like a scan of the Roche "Biochemical Pathways" poster, but with the Roche logo and some text removed from the upper left corner. I used to have the physical poster; it's a very different experience studying the giant poster versus zooming in with a browser.
If you view it in a single graph... yes, it seems completely intractable. If you can break it down into subcomponents and track the dependencies, it becomes much more tractable.
The thing with biochemistry is that it is highly connected, and very complex. Molecules aren't just products, they can also bind to gene regulators, or stabilize protein complexes. You can't cleanly separate parts of the system.
Reminds me of a talk by Andrew Endy[1], a long time ago, about synthetic biology. The essence is that his vision is to create "building blocks" of bio that have standard interfaces so you can then "code" them instead of untangling this spaghetti of life.
Typical case. The original developer has abandonded the project, maintainers make a complete mess of it, project ends up electing Trump as president of the United States.
There have been several attempts to virtualize cells, and come up with expressive PLs for hacking metabolic pathways. Turns out shit is too complicated
http://www.reactome.org/PathwayBrowser