Supplementary Figure 10: Comparison between linear correlations and nonlinear decoding; effect of PP threshold
From: Thalamic nuclei convey diverse contextual information to layer 1 of visual cortex
(a) Relationship between the Pearson’s correlation coefficient R between inferred spike rate and running or visual flow speed, and the prediction power (PP) for speed traces obtained with the non-linear regression decoder for individual boutons. Black: boutons above the threshold of significant prediction power PP > 0.16. (b) Relationship between Pearson’s correlation coefficient R for running speed and visual flow speed for all LP (left) and dLGN boutons (right). Magenta (LP) and green (dLGN) boutons have a decoding |PP| > 0.16. Equivalent to Fig. 6d. (c) Circular histogram depicting the distribution of linear interaction angles θlin for running and visual flow speed computed from corr. coeffs of running and visual flow speed, for boutons with decoding |PP| > 0.16. Equivalent to Fig. 6e. Proportions of ±(RS−VF) boutons (including angles of up to 22.5 deg off the diagonal): dLGN 19%, LP 39%; P < 10−10, Z-test. Proportions of ±(RS+VF) boutons: dLGN 36%, LP 25%; P < 10−10, Z-test. dLGN: n = 2159 boutons; LP: n = 1617 boutons. (d) Scatter plots of signed decoder PP for running speed versus visual flow speed for all LP and dLGN boutons. Similar to Fig. 6d, colors indicate for which variable (running speed (RS), visual flow (VF), the difference between RS and VF (RS−VF) or the sum of RS and VF (RS+VF) the boutons had the highest prediction power. Grey: PP < 0.16 for all variables. LP: 10 mice, dLGN: 8 mice. (e) Circular interaction histograms showing the distribution of visual flow and running speed interaction angles computed from signed PPs, similar to Fig. 6e for boutons above varying PP thresholds. Proportions of ±(RS−VF) and ±(RS+VF) boutons were significantly different between dLGN and LP for all PP thresholds (P-values < 10−10, Z-test).
