Fig. 2: MLIs developed divergent responses during go–no go learning.
From: Molecular layer interneurons in the cerebellum encode for valence in associative learning
a Learning curve for a mouse discriminating 1% Iso from MO in the go–no go task. Magenta dots: proficient level (≥80% percent correct), green dots: naive level (percent correct ≤ 65%). n = 20 trials within a sliding window. b–d GCaMP6f fluorescence (ΔF/F) time course averaged (mean ± 95% CIs) over all ROIs and trials falling within different proficiency windows shown for the session whose performance is shown in (a). b ≤65% (naive, 21 trials), c 65–80% (12 trials) and d ≥80% (proficient, 30 trials), 167 ROIs. Orange: S+, light blue: S− (the shaded area is the 95% CIs). The vertical black lines are odorant onset and removal and the red lines bound the reinforcement period. e, f Per trial ΔF/F averaged (mean ± 95% CIs) over all ROIs for 1 s before odorant application (e Pre-odorant) and in the last second during odorant application (f Odorant) shown for the session in (a). n = 1 mouse, 167 ROIs. g Violin plot showing per ROI ΔF/Fs for the session shown in (a) for the following time windows: pre-odorant (1 s before odorant application), odorant (the last second of odorant application) and reinforcement (one and a half seconds after onset of reinforcement). Per ROI ΔF/Fs are shown for trials falling within different proficiency windows: ≤65% (naive) and ≥80% (proficient). A GLM indicates that there are significant differences between S+ (orange) and S− (light blue), between time windows and between naive and proficient mice (p < 0.001, 5538 d.f., one mouse, *post-hoc ranksum/two-sided t test p value < pFDR = 0.043). h, i. ΔF/F for four mice averaged (mean ± 95% CIs) over all the ROIs for all trials falling within different proficiency windows. ΔF/F average was calculated for 1 s before odorant application (h) and for the last second during odorant application (i). For the data during odorant application (i) GLM analysis indicated that there is a statistical significance for the interaction between percent correct and the identity of the odorant (p < 0.001, 43 d.f., n = 4 sessions, 4 mice, GLM F-statistic = 60, p < 0.001). GLM did not yield statistically significant differences for pre-odorant data (h p > 0.05, 43 d.f., n = 4 sessions, 4 mice, GLM F-statistic = 0.46, p > 0.05). *Post-hoc two-sided t test, p < pFDR = 0.018. Error bars are 95% CIs.
