Question: How does Cufflinks DE calculate q-value?
0
gravatar for john.perish
17 months ago by
john.perish0 wrote:

I have run my aligned samples through the Cufflinks app in Illumina BaseSpace a couple of different ways. When I do Classic FPKM Normalization + Pairwise comparisons, I get a small list of genes. This analysis seems stringent. When I do Classic FPKM Normalization + No Pairwise comparison, I get a very long list of genes significantly upregulated when I do t-tests on my own. But there is very little overlap between the two lists. If the Pairwise expression tests are more stringent than a t-test of FPKM values, most of the pairwise comparison genes should show up on the t-test list, right?

So my question is: How exactly is the Test statistic, p-value, and q-value calculated on the Pairwise comparisons in Cufflinks Differential Expression? I’m trying to re-create this statistical analysis on my raw FPKM value datasets but I have not been able to.

ADD COMMENTlink modified 12 months ago by kristoffer.vittingseerup1.7k • written 17 months ago by john.perish0
1
gravatar for Hussain Ather
17 months ago by
Hussain Ather910
National Institutes of Health, Bethesda, MD
Hussain Ather910 wrote:

Check out the Benjamini-hochberg procedure wikipedia page.

And this paper explains the p-value and q-value calculations.

ADD COMMENTlink written 17 months ago by Hussain Ather910
0
gravatar for kristoffer.vittingseerup
17 months ago by
European Union
kristoffer.vittingseerup1.7k wrote:

If you want to recalculate anything you should do it based on the (estimated) count values since they contain the read-count uncertainty (much higher for 3 reads than 300 reads) - somthing you normalize away when you calculate FPKM.

ADD COMMENTlink written 17 months ago by kristoffer.vittingseerup1.7k
0
gravatar for aqabulel1977
13 months ago by
aqabulel19770 wrote:

Advertising is your main tool to have the business ready to go. Actually, everybody is busy in their company marketing over the World Wide Web. Unfortunately, online company is scammed with scammers. On occasion the provider hires you freelance, or occasionally it is a site designer which orders the copy google At an identical time, if you decide on the most suitable company having excellent reviews and reputation, you will enjoy well composed works. http://google.com/

ADD COMMENTlink written 13 months ago by aqabulel19770
0
gravatar for isabellebaldwin99
13 months ago by
isabellebaldwin990 wrote:

I don't know. I hated the statistics when I was studying in the college. I picked Psychology because I thought there won't be any math. i've never been so wrong. It has a lot of math in statistics. At least I had big discounts by Bid4Papers.com,s o I just used their help each time, when math was involved.

ADD COMMENTlink written 13 months ago by isabellebaldwin990
0
gravatar for kristoffer.vittingseerup
12 months ago by
European Union
kristoffer.vittingseerup1.7k wrote:

That would be very hard - Cufflinks/Cuffdiff uses a quite advanced statistical model. For more information take a look at this article - the supplementary methods contains the mathematical description of the model.

There are other better methods if you only have FPKM values, but do you have access to the corresponding count data?

ADD COMMENTlink written 12 months ago by kristoffer.vittingseerup1.7k
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 2.3.0
Traffic: 1783 users visited in the last hour