The development of therapeutics depends on predictions of clinical activity from pre-clinical data. We have previously described SYNB1618, an engineered bacterial therapeutic (synthetic biotic) for the treatment of Phenylketonuria (PKU), a rare genetic disease that leads to accumulation of plasma phenylalanine (Phe) and severe neurological complications. SYNB1618 consumes Phe in preclinical models, healthy human volunteers, and PKU patients. However, it remains unclear to what extent Phe consumption by SYNB1618 in the gastrointestinal tract lowers plasma Phe levels in PKU patients. Here, we construct a mechanistic model that predicts SYNB1618 function in non-human primates and healthy subjects by combining in vitro simulations and prior knowledge of human physiology. In addition, we extend a model of plasma Phe kinetics in PKU patients, in order to estimate plasma Phe lowering by SYNB1618. This approach provides a framework that can be used more broadly to define the therapeutic potential of synthetic biotics.