Molecular Nearest Neighbors Determine Generalization of Odorant Mixtures

摘要: Most natural odors arise from mixtures of multiple odorants. Some such are perceived "elementally", with each odorant component clearly identifiable, while others "configurally", the mixture adopting a perceptual quality distinct any components. While similarity two is presumably related in some way to corresponding components, given elemental/configural dichotomy it unclear if formal principle can be used predict similarity. To investigate this problem, we trained mice respond binary reference structurally similar odorants (S+) and then tested generalization response other test mixtures. Across 5 experiments, parametrically varied these ways candidate models for The best-performing model predicted behavioral responses by considering, S+, only most ("nearest neighbor") mixture. We conclude that tasks olfactory system may deemphasize or discard information about components not perceptually "near" enough those consistent sparse elemental rule perception structurally-related