A great example of the gap between our current models and what would be sufficient comes from the COX-2 inhibitors (Vioxx/Celebrex). Ibuprofin (Motrin/Advil) and other NSAIDs modulate both COX-1 and COX-2 receptors in the body to reduce inflammation, but inhibiting COX-1 also causes stomach damage. The early COX-2-specific inhibitors were designed from the ground up using our best models, which were thought to be well-understood. Unfortunately, after several years on the market, it was discovered that this also came with a large increased risk of heart attacks and strokes.
The body just has so so many moving pieces, and while we're getting better, it's going to be some time before we are able to base these studies entirely on theory, and when we get there we'll be able to skip the mice entirely!
So add up random biochemical pathway * X million years of evoltion, and one begins to understand the fun of drug discovery and testing.
"We know this is the pathway involved in heart disease. We know this chemical compound is effective against that pathway.
... Now why are all our mouse models getting erections?"
It's still very much a "we don't know what we don't know" situation, which is both intimidating and exciting. There's a lot to be done before our models are even adequate, but we at least get to be part of that learning process.
Correct. We understand bits and pieces of biology but we don't understand the entire system well enough to predict anything with any confidence.