Extended Data Fig. 9: Cell tracking across sessions.

a, Panels i–iv show the spatial footprints of different cell types in a representative session. bi, A zoomed-in view of a reward cell, with the green contour outlining its cell body. bii, Tracking of this cell across all sessions (24 sessions for this mouse). For each session, the top panel shows the spatial footprint with the cell body outlined in green. The bottom panel shows the across trials averaged calcium response relative to reward onset (mean ± s.e.m.). biii, Calcium responses across sessions show a backward shift. The left panel shows overlapping contours of cell bodies. Each contour is for one session. biv, The same analysis as in biii was repeated, but for each session we replaced registered cells with a nearby randomly selected cells. This shows that the backward shifting disappears. c, Similar to b, for another cell. d, We systematically varied the percentage of misaligned sessions for tracked reward cells. For each misalignment level, we ran 1,000 iterations, randomly selecting a percentage of sessions in each iteration. In these sessions, the registered cell was replaced with a randomly chosen neighbouring cell within a five-cell-diameter radius. As misalignment increased, the negative correlation between reward response and time weakened. e, The results for similar analysis as in d for the entire tracked reward cells. Bar graphs and error bars in d,e show mean ± s.e.m.