When deciding what to eat, animals evaluate sensory information about food quality alongside multiple ongoing internal states. How internal states interact to alter sensorimotor processing and shape decisions such as food choice remains poorly understood. Here we use pan-neuronal volumetric activity imaging in the brain of Drosophila melanogaster to investigate the neuronal basis of internal state-dependent nutrient appetites. We created a functional atlas of the ventral fly brain and find that metabolic state shapes sensorimotor processing across large sections of the neuropil. By contrast, reproductive state acts locally to define how sensory information is translated into feeding motor output. These two states thus synergistically modulate protein-specific food intake and food choice. Finally, using a novel computational strategy, we identify driver lines that label neurons innervating state-modulated brain regions and show that the newly identified ‘borboleta’ region is sufficient to direct food choice towards protein-rich food. We thus identify a generalizable principle by which distinct internal states are integrated to shape decision making and propose a strategy to uncover and functionally validate how internal states shape behaviour.