Begin with the physiological question
Aeviant starts by defining the dysregulated state and the useful function that should remain intact. Its current public scope is physiology-led small-molecule research; it does not claim modality neutrality or a validated platform.
Evaluate selectivity at more than one level
Target and subtype preference matter, but so do biological state, timing, signalling bias, endogenous demand and functional consequence. None of those properties follows automatically from a pharmacological label.
Let measured evidence guide the mechanism
Computational work can characterize a target, generate hypotheses and prioritize future experiments. It cannot show that a molecule binds or that the proposed pharmacology occurs in a living system. Those claims require experimental evidence.
Allow the program to change
The original belief is not protected from contradictory results. A program may advance, redirect or stop when predefined criteria fail or the evidence no longer supports the model.