how to understand the P value of the Deseq2 results

Question

[Kai_Qi]

Hi:

I am trying to get a better understanding of how the results of DEseq2 come out, I spend a few days to get to understand the p_value. I don't know if my understanding is right. So I posted it here for discussion:

Let's assume we have 4 samples control (S1, S2); Knockout (S3, S4)

1. the reads number(a value) of genes follows negative binomial distribution, so the reads in S1, S2, S3, S4 will each has a mean and variance (from the reads of each gene).
2. When doing the analysis, fold change, for example, Log(S3+S4)-Log(S1+S2), it will generate a new distribution (let name it S5) and this new distribution will have its own mean and variance.
3. Now I have a new of distribution with values represent the differences between the reads of each gene from control and knockout. And the distribution will have a mean and a variance.
4. use the mean and variance to evaluate the probability of each value in the new distribution (S5), if the P value < 0.05, we see that the differences of counts between control and knockout is significant, if not we accept the null hypothesis.

I am not sure my understanding on how Deseq2 generate p value is right or not. Thanks ahead for anyone that comments on this.

Best,

[Kevin Blighe]

Has been cross-posted: https://support.bioconductor.org/p/9136691/

[Kai_Qi]

Sorry for not mention earlier. I mentioned it when posted it on bioconductor. I would like to get some mentoring on the principles. I have some problem in connecting book with practical questions.

Thanks for pointing it out

[swbarnes2]

Did you try to read the DESeq paper?

[Kai_Qi]

Read the tutorial and the paper. Tutorial is relatively easier, whereas the original paper is a little bit harder. So I turned to my college books(in my native language), and tried to understand better. Thanks, I will keep reading the original paper

Answer 1

[Michael Love]

There is this previous post:

Can someone please explain in simple terms how DESeq2 works?