Open laufre opened 3 years ago
Hey @votti I also wondered where the recommended median pixel threshold of 200 counts in your scripts was coming from. Did you ever test how changing the pixel bin for aggregation affects the results? Or would you just recommend to bin to a point where you get a median of > 200 counts per spot?
Hi,
I'm wondering on how to correctly use the data bining in imc_generatespillmat_long.
# defines over how many pixels the aggregation should happen # 1 = no aggregation npixelbin = 1 dats_agg <- lapply(dats_raw, function(x) aggregate_pixels(x, n=npixelbin)) dats_agg_sum = rbindlist(lapply(dats_agg, calc_file_medians), idcol = T)
Our median per-pixel Intensities are quite low, compared to your example dataset (Sth. like 1:100 lower) and without data bining the generated matrix displays spillover only between very few channels. I think, that this spillover matrix is not correct, as it is totally missing spillover between channels like 172Yb and 173Yb for example. If I increase
npixelbin = 1
to e.g. =2, =10 or =50 the spillover values change quite a bit, looking more like what to expect. Nevertheless I'm now wondering, how to use this feature correctly, without accumulating too much background noise, resulting in wrong spillover values. How do I pick the correct value for data bining?Thank you in advance for an explanation!