meyer-lab/DDMC

FigureM2

Closed this issue · 4 comments

mcreixell commented

This is what this figure currently looks like:

figureM2

These are items I was thinking would make this figure final:

  • Add error to minimum peptide signal in addition to average
  • Compare to other imputation strategies: GMSimpute / DreamAI
  • Add legends in plots B-F.

What do you think? @aarmey

aarmey commented

Makes sense. What prevents you from going higher than 22 clusters?

mcreixell commented

24 is the highest number of clusters with the binomial method, 22 with pam250. More than this we get nans in the scores and breaks via an assertion.

aarmey commented

Hmm... can you tag me in an issue saying that? I'd like to see why.

mcreixell commented

Do you think that the improvement of DDMC vs k-means/GMM seen in Supplemental Figure 5 (#334) is enough to skip the imputation evaluation?