Fig. 5: Identification of stimulus-, choice-, reward- and motor-related activity patterns of single neurons using a generalized linear model.
From: Mouse visual cortex areas represent perceptual and semantic features of learned visual categories
a, Left, example section of a design matrix. The y axis shows regressor sets (cat., category; spat. fr., spatial frequency; rew., reward, T.O., time-out; Supplementary Table 3), and the x axis shows time and behavioral trials. A regressor set consists of multiple rows with time-shifted gaussian curves (see b). Right, fitted response kernels of an example neuron. b, Example regressor set with multiple rows of gaussian shaped curves (left). Nonnegative least squares (NNLS) fitting using L1 regularization estimated the weight of each curve (right). c, Inferred spike activity (gray represents data, and purple denotes the model prediction) of a single neuron for an entire imaging session. The below inset shows a zoom-in view of the model-fit that captures the amplitude and timing of inferred spiking activity. d, Stimulus-triggered inferred spike response of an example neuron (area POR, mouse M16) to all ten category stimuli organized in a grid of spatial frequency (y axis) by orientation (x axis). Solid lines in light blue represent the left category, and pink lines represent the right category. Vertical scale bar, 0.1 inferred spikes; horizontal scale bar, 2 s. The dashed black lines represent spatial frequency kernels, while solid black lines denote orientation kernels. The solid gray lines represent choice kernels. Dark blue/red solid lines are category kernels. Left, time point ‘baseline 3’; right, ‘learned 1’. e, Classification of neurons based on the significance of their modulation in the in-task time points ‘baseline 3’ and ‘learned 1’. f, Mean (±s.e.m.) fraction of changed (‘gained’ and ‘lost’) neurons divided by the fraction of ‘stable’ neurons in dorsal and ventral stream-associated areas. Gray dots denote individual chronic recordings (two-sided Mann–Whitney U test, U = 41, P = 0.009; ndorsal = 12, nventral = 15 chronic recordings from ten mice). g, The fraction of neurons that showed significant unique modulation (ΔR2) by the group of regressor sets (y axis), separately per imaging area (x axis), for the groups defined in e. Right, difference in the fraction of neurons that showed significant unique modulation in time points ‘baseline 3’ and ‘learned 1’. White asterisks denote a significant difference from zero (chi-squared test, P values were calculated using Bonferroni correction for 63 comparisons). NS, P > 0.05, *P < 0.05, **P < 0.01.
