Question: Handling multiple knockout experiments with DESeq2
0
gravatar for cmfield
10 months ago by
cmfield0
cmfield0 wrote:

This is a sanity-check question about how I ought to correctly use DESeq2 with knockout experiments. If I have two conditions, a control and a knockout, am I correct in putting them into one data set and then later ignoring the specific result for the knocked-out gene? I would guess that should a normally high-count gene be knocked out, the subsequent scaling up of every other gene is accounted for by the analysis, but that genes previously unseen (zero count) that are then seen in the knockout condition may have their significance mis-estimated (under-estimated?) due to uncertainty about their true frequency when unseen.. but would welcome clarification on this.

If instead I have multiple conditions with different genes knocked out should they all be put together for normalisation/analysis, or should I construct sub-tables with the control and each knockout separately?

deseq2 • 249 views
ADD COMMENTlink modified 10 months ago by Michael Love26k • written 10 months ago by cmfield0
Answer: Handling multiple knockout experiments with DESeq2
1
gravatar for Michael Love
10 months ago by
Michael Love26k
United States
Michael Love26k wrote:

How many genes are expressed in control and knock-out, roughly? E.g. how many genes with counts > 10 for samples in these two conditions? The size factor estimation uses the median of ratios over all genes so it is certainly robust to a single gene, and is typically robust to there being many DE genes, as long as the center of the distribution of log ratios is not shifted: e.g. not all genes up-regulated or down-regulated.

ADD COMMENTlink written 10 months ago by Michael Love26k

On the order of 100 I would say, not so many. I suppose another way to look at it is.. is it worth separating the knockouts into different analyses at the cost of less robust size factor estimation?

ADD REPLYlink written 10 months ago by cmfield0
1

Sorry, I meant: how many genes are expressed in control? How many genes are expressed in knockout? And how many genes expressed in both?

ADD REPLYlink written 10 months ago by Michael Love26k

It's approximately the same set in each condition (other than some that go from <10 to >10 in a knockout), of 50-100.

ADD REPLYlink written 10 months ago by cmfield0
1

I'd guess you have more genes expressed, and what you are telling me is the intersection?

Not knowing further details, I would say to construct sub-tables of control and each knockout separately.

ADD REPLYlink written 10 months ago by Michael Love26k

Well, to be precise, I'm saying 'genes' when I mean 'countable things' because it's not RNASeq data, which is one of the reasons why I have very low numbers of countable things > 10. I didn't want to complicate the original post but more understand the theory. I have so far constructed sub-tables because I figured there were too many knockouts relative to the number of different countable things and I guess I'll stick with that.

Thanks so much for the replies, I am amazed that you manage to remain so active in helping with the package!

ADD REPLYlink written 10 months ago by cmfield0
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 398 users visited in the last hour