smallest number of cells per bin

[Daniel Gingerich]

Hi, 

when running create_cicero_cds, I was wondering if there was a minimum k value (cells per bin) that you recommend.  I am running cicero on individual clusters, and k = 50 is quite large considering the size of my clusters.  My cluster sizes are:

Astro1   Exc1  Exc10  Exc11   Exc2   Exc3   Exc4   Exc5   Exc6   Exc7   Exc8
  4725   5815    337    153   2455   1686   1495   1262   1205    506    471
  Exc9   Inh1   Inh2   Inh3   Inh4   Inh5   Inh6 Micro1 Micro2 Oligo1 Oligo3
   395   1927   2062   1651   1374   1198    341   5804    245  20099   4945
Oligo4 Oligo5 Oligo6   OPC1
  5344   5658   5150   3468

  Would k = 10 be a reasonable size?  

[hpl...@gmail.com]

Hi,

The idea of the aggregation is to move out of a basically binary regime. If your k is too small, then for the majority of sites, you'll still basically be binary... it will depend somewhat on the efficiency/depth of your libraries. Intuitively k of 10 sounds low to me... maybe try 20 and see how it goes?

Best,

Hannah

[Daniel Gingerich]

Hey Hannah, 

Thanks! That is reassuring to hear, because I had the same thought.  I ended up using k = 20.  results seem OK to me but still need to give it the comprehensive check