How does regularized log transformed data from DESeq2 relate to read counts?

Question: How does regularized log transformed data from DESeq2 relate to read counts?

2.8 years ago by

Hi,

I have used DESeq2 to analyse some RNAseq data and am using regularized log transformed data (rlog) output from DESeq2 to plot heatmaps.

However I am struggling to understand how the rlog output relates to the actual number of reads obtained for that gene. Can anyone explain this to me? For example, if I have a regularized log value of -1 for a particular gene, does this mean that I had less than 1 read for that particular gene in my sample (if so, how can this be)?

The reason I would like to understand this is I want to set a threshold for gene expression, so I am only looking at well expressed genes in my analysis. Is this possible with regularized log transformed data? Or would it be better to use a metric like RPKM? I was under the impression that working with regularized log transformed data is more advisable than RPKM, but perhaps in this instance RPKM is more appropriate.

Many thanks in advance for your help, Amy

rna-seq R • 3.5k views

ADD COMMENT • link •

modified 2.8 years ago by tarek.mohamed • 260 • written 2.8 years ago by amymccarthy91 • 10

2.8 years ago by

tarek.mohamed • 260

tarek.mohamed • 260 wrote:

Hi rlog function produces log2 scale which has been normalized with respect to library size. So for example a P-value of 0.05 is -4.3 on the log2 scale, and P-value of 0.005 is -7.6. gene expression of 100 read counts will be 6.6 on log2 scale.

Tarek

ADD COMMENT • link modified 2.8 years ago • written 2.8 years ago by tarek.mohamed • 260

It's not the P-values that are on the log2 scale, it's the number of counts.

ADD REPLY • link written 2.8 years ago by mastal511 • 2.0k

I am just giving an example by the P-value! you can replace the word "P-value" with anything! gene expression of 100 read counts will be 6.6 on log2 scale.

ADD REPLY • link modified 2.8 years ago • written 2.8 years ago by tarek.mohamed • 260

Hi Tarek, Thanks for your answer. So is the rlog is essentially log2(counts/library size)? If so, how can I have positive rlog values (since the number of read counts must always be less than the library size?) Amy

ADD REPLY • link written 2.8 years ago by amymccarthy91 • 10

Hi Amy, DESeq2 performs normalization where geometric mean is calculated for each gene across all samples. The counts for a gene in each sample is then divided by this mean. The median of these ratios in a sample is the size factor for that sample.

Tarek

ADD REPLY • link written 2.8 years ago by tarek.mohamed • 260

Hi Tarek, Okay, that helps explain it. Thanks for your help! Cheers, Amy

ADD REPLY • link written 2.8 years ago by amymccarthy91 • 10

Similar posts • Search »