search

navinlabcode / copykat

Other
213 stars 55 forks source link

Issue recapitulating results for IDC1 #55

Closed inofechm closed 2 years ago

inofechm commented 2 years ago

Hi thank you for the tool it is great. I am having some inconsistencies with recapitulating your results (figure 4 from the paper for IDC1) here is my code, the only tweak I made was some filtering of low quality cells but this shouldn't make a difference. Additionally I am using a sparse matrix.

test <- copykat(rawmat = rna_counts, n.cores = 4, sam.name = "IDC1") table(test$prediction[,"copykat.pred"])

aneuploid diploid 1511 1275

My results are not splitting the populations as they are in the paper and even finding a large population that seems to be CNV low? could you please clarify if there are any default settings that were changed when you generated the heatmap? IDC1_copykat_heatmap

gaobio commented 2 years ago

@inofechm Yes, I recalled that sample was very challenging, with aneuploidy around 2 (I'd double-check). There is a different distance method other than the default. Also it also need a smaller LOW.DR. In the cases of low aneuploidy, I had to tune several parameters.

inofechm commented 2 years ago

Thank you for the response! I was wondering if you could provide any details on how one knows when to switch these settings, if they do not how much cancer vs normal cells are present. I appreciate that this was a tricky sample, but how do I know whether a sample is tricky or whether it has lower tumour cellularity and lots of normal cells?