Animal testing is not going away in drug discovery. Not because animal models are perfect proxies for humans — they obviously are not. It is because a drug has to enter a complex adaptive system before it enters a human one.

The reductionist replacement story

A lot of the rhetoric around replacing animal studies assumes drug discovery is mainly a reductionist prediction problem: predict binding, predict permeability, predict clearance, predict tox, predict efficacy from the right in vitro setup, and eventually the whole organism becomes redundant.

That framing misses the real issue. In vivo biology is not just a bigger assay. It is a web of feedback loops, compensatory responses, tissue effects, metabolite formation, exposure mismatches, immune interactions, and failure modes you did not think to look for.

The problem is not model accuracy

The problem is not merely that our models are imperfect. It is that the space of possible organism-level interactions is too large to fully specify in advance. We cannot even reliably predict something as apparently narrow as small molecule + protein binding affinity across the board. The idea that we will soon model the full downstream consequences of delivering that molecule into a living organism is fantasy.

Animal studies persist because they are the first place where biology gets to answer back as a system. Not a perfect system. Not a human system. But a system — one where emergent interactions, compensatory pathways, and off-target surprises actually get a chance to show up before a human is exposed.

What the serious goal looks like

The serious goal is not “animal testing disappears.” It is: use better in vitro models, better computation, and better decision-making so that in vivo studies happen later, with fewer compounds, clearer hypotheses, and much higher information content per study.

That is a real improvement. Pretending whole-organism uncertainty has been solved is not.