How a model earns a place in discovery

Discovery programs rarely run out of compounds. They run out of months. When every candidate looks equally worth a screen, the scarce resource is time in the lab, not ideas on a slide.

We use models to organize what is already known: structure, predicted properties, and how one candidate compares with the rest of the set. Something moves up. Something waits. The team can say why. That ranking is a priority list, not a drug.

Any prediction the lab cannot check is a suggestion, not a plan. We mark where the model is thin so the experiment is designed around the uncertainty instead of hidden by it. An honest short list saves more calendar time than an optimistic long one.

If you are deciding whether computation belongs in a program, pick one choice the team already struggles to make and narrow only that. Widen the use of the model after the shorter list has earned its place at the bench.