Extended Data Fig. 2: Computational models for predicting calcium signals in neurons.

a, Activation functions for claw, hook flexion, hook extension, and club and 9A neurons. b, Measured and fitted calcium signals of claw axons in response to applied ramp-and-hold movements of the femur-tibia joint (N = 10 flies per ramp-and-hold stimulus). Lines, mean of animal means; shadings, s.e.m. c, Cross-correlation between measured and fitted calcium signals at a time lag of zero (N = 10 flies, n = 20 trials in total). Black line, median. d, Same as b but for hook flexion axons (N = 14 flies). e, Same as c but for hook flexion axons (N = 14 flies, n = 28 trials in total). f, Same as b but for club axons (N = 14 flies). g, Same as c but for club axons (N = 14 flies, n = 28 trials in total). h, Response of the hook flexion model to an artificial femur-tibia joint angle without tracking noise. When the velocity threshold is larger than −50 deg/s, the model starts to predict strong calcium signals in response to slow flexions, as seen in the measured calcium signals. In b–g, experimental data are from ref. ²¹.