Abstract
The circuits of olfactory signaling are reminiscent of complex computational devices. The olfactory receptor code, which represents the responses of receptors elicited by olfactory stimuli, is effectively an input code for the neural computation of odor sensing. Here, analyzing a recent dataset of the responses of human olfactory receptors (ORs) to odorants, we show that the space of human olfactory receptor codes is partitioned into a modular structure where groups of receptors are “labeled” for key olfactory features. The existence of such independent and sizable receptor groups implies a significant dimensional reduction in the space of human odor perception. Our data-driven statistical analysis of receptor codes leads to a valuable discovery that human olfaction works by hybridizing both the combinatorial coding and labeled line strategies, even at the early stage of signal processing.