search

abelson-lab / scATOMIC

Pan-Cancer Single Cell Classifier
MIT License
58 stars 5 forks source link

Problem too large #30

Open YaobangWang opened 4 months ago

YaobangWang commented 4 months ago

Hi, I'm glad that scATOMIC helped me a lot in distinguishing tumor from non-tumor cells. However, when I ran a large dataset of 67000 cells, the following error occurred: "Cholmod error 'problem too large' at file ../Core/cholmod_dense.c, line 105". This large dataset is a combination of single-cell RNA sequencing data from 25 cancer patients. I wonder how I should do the right analysis in this situation? Plan A: I will divide this large dataset into 25 sub-datasets on a per-patient basis, and then analyze each sub-dataset individually. Plan B: I will split these 67,000 cells into two equal parts at a 1:1 ratio, and then analyze each part separately. Which of the above two plans is the better choice? I would be very grateful, if you have any suggestion to solve this problem!

inofechm commented 4 months ago

Hi Yaobang,

Thank you for your interest in scATOMIC! scATOMIC is designed and has assumptions built into the methods that there is only one tumour sample when it is run. So I definitely recommend trying Plan A for annotation, followed by merging all the annotation results. Please see issue #21 for some more details.

Best, Ido