dcjones
Splitting original cells along grid lines
Hi @dcjones
Thanks for developing Proseg. I am finding it is an improvement over Xenium default segmentation in my UMAPs.
I am using Proseg v2.0.5 to analyse Xenium data. I have used the original_cell_id to view cell groups in Xenium Explorer. I found that 0.7% of the original_cell_ids appear more than once in the Proseg output, which I understand are cells that have been split by Proseg. I exported the split cells (those cells with an duplicate original_cell_id) into Xenium Explorer and found that the split cells lie along grid lines. Presumably this is due to how Proseg batches the data for processing.
I think this will have the effect of double counting the cells that have been erroneously split, so as a work around I am thinking that I should merge cells with the same original_cell_id as the first step in my analysis? There don't appear to be many cells off the grid lines that have been split.
Thanks, Allan
Ah this looks like a bug. Proseg was assuming the same cell id doesn't occur in multiple fovs, which was to work around cosmx assigning non-unique cell ids. That seems to cause this error in xenium which has better stitching. I'll release a fix soon.
