Having explored (to a small extent) what we can do with Xenium data, we hope to dig deeper into biological insights. We look forward to combining this with the growing suite of available, collaborative tools for analyzing ST data.

Our code is available here: github.com/jmkotah/xeni....

We did this using a relatively crude threshold segmentation of IBA1 fluorescent signals to selectively allocate neuroimmune transcripts. Importantly, we expect this to be compatible with more advanced segmentation methods, which can be used to maximize the data obtained from imaging ST data.

Because the Xenium platform nicely allows for post-run immunostainings, we showed proofs-of-concept that registering and segmenting post-run images can be used to direct transcript allocation to cell types of interest, and that 'leftover' transcripts outside masks can be re-allocated.

We used the tools provided by geopandas to try different ways of allocating transcripts, and found that simply expanding the nuclear borders (as determined by DAPI) can have consequences, e.g., for cell type annotations. On the other hand, the strict nuclear border leaves a lot of data on the table.