Figure 7.

Adaptive properties of ab3A response quantified using an LN model. a, Top to bottom, On–Off state of the valve driving the odor; PID response measured at the fly; and raster plot and PSTH of ab3A response (black, n = 18) and prediction from LN model obtained using the filter shown in d (red). Stimulus sequence was 30 s long with 30 ms correlation time. b–d, LN models for valve-to-odor, valve-to-ab3A, and odor-to-ab3A transformations. Left to right, Linear filter k(t) and nonlinear static transformation N. NR indicates the quality of the model prediction, and a value <1 means that the prediction is within the variance of the response (see Materials and Methods). e, Adaptive response to flickering stimuli with different correlation times. All three correlation times were tested on each neuron (n = 13), and the stimulus sequences was 30 s long. Left, Valve-to-odor transformation is independent from the stimulus correlation time. Right, Odor-to-ab3A transformation shows a faster negative lobe for shorter correlation times. f, After the onset of a stationary flickering stimulus, ab3A and pb1A adapt their gain but not their response kinetics. Filters and static functions were extracted for the first 10 s (early response) and the last 10 s (late response) of a 60-s-long flickering sequence with 100 ms correlation time. Top, Linear filters are identical for early and late responses for all odor–receptor combinations. Bottom, The static function shows a decrease in slope of different degree for the different odor–receptor combinations.