Extended Data Figure 1-1
To evaluate our within cell statistical comparisons to identify “responsive” versus “non-responsive” neurons, in particular to test the possibility that drift might contribute to some of our identified drug effects, we conducted a sliding window analysis on a subset of our drug responses (all TH-positive neurons tested with bath application of 10 nm N/OFQ). Further, any increase in statistically significant sliding windows during drug washout compared to the static baseline would suggest underlying Iholding drift. We compared all 4-min windows from predrug application through drug washout to a fixed “baseline” window (the 4 min preceding the onset of the drug). To create the windows, Iholding of each recording was binned into 30-sec intervals and assigned a bin number (1, 2, 3 Ö n). The baseline eight-bin (4-min) window was compared with the target 4-min window by way of a Student’s unpaired t test. The p value and significance of the comparison were then corrected using the Bonferroni method for multiple comparisons. The alignment of the sliding window was then increased by a single bin and the comparison repeated, resulting in an array that represents all significant 4-min intervals for each drug effect. The resulting arrays were plotted as a histogram representing, at the initial bin time of the sliding window, the proportion of recordings in which this calculation was significantly different from the fixed baseline target window (A). In the neurons previously classified as responsive by a single baseline compared to “drug” window comparison, the rising left edge of the histogram begins to plateau around the 4th minute of drug application, consistent with the plateau of the mean effects across all cells reported in Figure 1D. Further, consistent with washout reversal of N/OFQ effects in most but not all neurons, the proportion of significant bins falls off as soon as N/OFQ application was terminated. That both the rise and fall of the frequencies of significant windows are time locked to the drug application suggests the response classification scheme is reliable. In neurons previously classified as non-responsive only one neuron had any significant windows, with three sliding window locations where this analysis yielded p < 0.05, suggesting that there was not systematic drift in these non-responsive neurons. In addition, a scatter plot (B) indicates the maximum number of consecutive significant sliding windows for each cell analyzed, because a well-behaved change in Iholding in response to the drug application should be detected in consecutive sliding windows. This graph shows that 8/12 neurons that were classified as responsive have more consecutive sliding windows different from baseline than the maximum found in non-responsive neurons. This analysis was conducted using a custom script created in Python (available at https://osf.io/c8gu7/?view_only=24595243ef6d44d5974442b23dda0b1d). Download Figure 1-1, EPS file.