seth-ament
commented
4 years ago Looks very nice, Brian! Just what I'd hoped we could produce for the MERFISH data. I agree that dots (not cell shapes) at the appropriate x,y position is just fine. However, I do wonder if the interpretation would it be made easier by producing drawings that simply show the inferred boundaries between layers and other major subdivisions.
The other visualization that I could potentially imagine creating would be an "imputed" version, integrating data from MERFISH and scRNA-seq. Specifically, I'm trying to think how one could get the MERFISH profile to show information about all the genes, not just the 250 that they imaged. Since the cells in the MERFISH and 10x datasets are given the same cell type assignments, would it be useful to assign each cell an expression level for every gene based on the median value for that cell type in the 10x data?
carlocolantuoni
brianherb
hertzron
jorvis
Merfish - Looking at the data, we actually have enough to make good looking plots. The actual experiment involved 2 mice - and they imaged 30 coronal slices per mouse (probing for 258 genes). For each slice, we have the xy position, size and cell type for each cell. The attached plots reproduce flagship figure 3d which looks at just the IT neurons (flagship figure: https://drive.google.com/drive/folders/1fFM1AnKbq0o0PXRZTRxlqiZuOS7g_YTt.) Each plot is a different slice and I sized the points on the plot relative to the cell volume. There is also a corresponding counts matrix for the ~300,000 cells and 254 genes, but I am having difficulty matching the cell ids.
The major difference is that we don't have the actual images, so the cells are depicted as circles instead of true cell outlines, but I feel like this is good enough, and might make plotting a lot easier. (seems like a lot of work to produce 60 SVG's too)