10Transformations between sensory representations are shaped by neural mechanisms at the cellular 11 and the circuit level. In the insect olfactory system encoding of odor information undergoes a 12 transition from a dense spatio-temporal population code in the antennal lobe to a sparse code in 13 the mushroom body. However, the exact mechanisms shaping odor representations and their role 14 in sensory processing are incompletely identified. Here, we investigate the transformation from 15 dense to sparse odor representations in a spiking model of the insect olfactory system, focusing 16 on two ubiquitous neural mechanisms: spike-frequency adaptation at the cellular level and lateral 17 inhibition at the circuit level. We find that cellular adaptation is essential for sparse representations 18 in time (temporal sparseness), while lateral inhibition regulates sparseness in the neuronal space 19 (population sparseness). The interplay of both mechanisms shapes dynamical odor representations, 20 which are optimized for discrimination of odors during stimulus onset and offset. In addition, we 21 find that odor identity is stored on a prolonged time scale in the adaptation levels but not in the 22 spiking activity of the principal cells of the mushroom body, providing a testable hypothesis for the 23 location of the so-called odor trace.
24Central to our modeling approach are two fundamental mechanisms of neural computation 51 that are ubiquitous in the nervous systems of invertebrates and vertebrates. Spike-frequency 52 adaptation (SFA) is a cellular mechanism that has been suggested to support efficient and 53 sparse coding and to reduce variability of sensory representation [21, 22, 23]. Lateral inhi-54 bition is a basic circuit design principle that exists in different sensory systems, mediates 55 contrast enhancement and facilitates stimulus discrimination [24, 25, 26, 27]. Both mech-56 anisms are evident in the insect olfactory system. Responses of olfactory receptor neurons 57 (ORNs), local interneurons (LNs) and PNs in the antennal lobe (AL) show stimulus adap-58 tation [28, 29] and strong adaptation currents have been identified in KCs [30, 31]. Lateral 59 inhibition in the AL is mediated by inhibitory LNs [32]. It is crucial for establishing the 60 2 Results 3 population code at the level of PNs [29, 33], for gain control [34, 35], for decorrelation of 61 odor representations [36], and for mixture interactions [29, 37, 38]. 62Taken together, we find that lateral inhibition and spike-frequency adaptation account for 63 the transformation from a dense to sparse coding, decorrelate odor representations, and 64 facilitate precise temporal responses on short and long time scales. 65