I was wondering (1) how does
( )calculate values from raw counts, (2) the relationship between rlog counts and the effect estimator (βir), and (3) what information rlog counts tell.
- log2(qij) = Xj βi
Q1: Is log2(qij) here represent the values from
Q2: My understanding from DESeq2 paper is that log2(qij) are the values calculated from Xj βi. Is that correct?
Q3: If Q2 is yes. What's "log qij" values for log qij=∑r Xjr βir? My understanding is: log (raw counts/size factor) instead of log2. Is that correct?
I saw many research made a heatmap of rlog counts. But I'm not sure what kind of information can rlog counts can tell me.
- Counts comparisons from my analysis are:
Control_1 Control_2 Control_3 Trt_1 Trt_2 Trt_3 Raw count 135 258 190 0 0 0
count(DESeq, normalized = T)
151.73 249.95 198.23 0 0 0 rlog(object, blind=F) 6.50 6.87 6.69 5.33 5.34 5.35 rlog(object, blind=T) 6.50 6.89 6.70 5.20 5.21 5.22 Size factor 0.89 1.03 0.96 1.17 1.02 0.98
- And the log2fold change are:
result(object) lfcShrink(res, type = "normal") lfcShrink(res, type = "apeglm") -10.16 -2.68 -11.24
It seems either
count(DESeq, normalized = T)or log2foldchange can illustrate the result better. Why we use rlog count for count heatmap?
Would you recommend to present the result by using rlog count heatmap? or log2foldchange?