search

dviraran / xCell

Cell types enrichment analysis
181 stars 61 forks source link

Some questions about xcell occur to me #37

Open huitingxiao opened 4 years ago

huitingxiao commented 4 years ago

Dear Dr.dviraran, Recently，I have tried my best effort to study the xcell methodology,unfortunately,there remains some questions bother me.I would appreciate if you have time to help me. 1.In the generate the signatures session,many cell types only have few samples(three or less),how to definite 10th, 25th, 33.3th, and 50th percentiles of low expression (Q1q), and 90th, 75th, 66.6th, and 50th quantiles of high expression (Q21-q). 2.Why to choose the minimum cell abundance is 0.8% ,instead of 0.1%，1% etc.

  1. When fit a power function，why shift the scores to 0 using the minimal score and divided by 5000? 4.The TCGA results of xCell providing Additional file 6,I found some scores are negative,these results how to interpret? I am sorry if these questions are a bit basic. Best wishes， huitingxiao
dviraran commented 4 years ago

Hi,

  1. Yup, the signatures will be similar. It doesn't really matter much, we are trying to create as many signatures as possible, and later choose those that show the most promise.
  2. No reason, it just the multiplication of 125.
  3. I think this explained in the paper. The ssgsea algorithm gives very different minimum scores and distribution based on the number of genes. After shifting the scores are all between 0 and ~4500.
  4. The negative numbers are all aprox 0 (i.e. -1e-16). I now round it to have 4 digits at most.

Hope this helps.

Best, Dvir

huitingxiao commented 4 years ago

It's very kind of you.Thanks for your patience and answer.