Question: Fake good model design: sum(result$adj.P.Val<0.05) [1] 1
0
gravatar for BioLite
7 months ago by
BioLite20
BioLite20 wrote:

Hi, friends, Is there someone could correct my default codes or thoughts? The highest appreciate from me. There are 6 samples with 2 kinds of treatments, like this:

array treatment GSM1 control-1 GSM2 control-2 GSM3 control-3 GSM4 Oxali-1 GSM5 Oxali-2 GSM6 Oxali-3

My aim is to compare the differential genes between "control" group and "Oxali" group, so I code like this:

> group=factor(rep(c("control","Oxali"),each=3),levels = c("Oxali","control"))

> group2=relevel(group,"control")

> design<-model.matrix(~factor(group2))

> rownames(design)=colnames(eSet)

> fit =lmFit(eSet,design)

> fit2=eBayes(fit)

> result=topTable(fit2,coef = 2,n=Inf,adjust.method ="BH",sort.by="P")

> sum(result$adj.P.Val<0.05)
[1] 1

The above codes look good, however, when I sum(result$adj.P.Val<0.05), there is only 1 result. That is so unreasonable, I must make some wrongs, while I don't find it. Another, the published article informs me that there are 267 up-regulation genes and 65 down-regulation genes, despite I set my cut-off values as p.value<0.05 & abs(logFC)>2, I get a wrong result different with article result.

My codes look good while working badly, I hope there are some kind friends can point out my drawback. Thank you very much!

ADD COMMENTlink modified 7 months ago by ATpoint15k • written 7 months ago by BioLite20

How did they analyze it in this paper? Any proper statistical framework behind it or just something like "fold-change > 1.5 in at least two of three samples"?

ADD REPLYlink modified 7 months ago • written 7 months ago by ATpoint15k

Thanks, ATpoint, I only can find this delineation——"The expression of 41,096 genes (Figure 3A) in liver tumor tissues from oxaliplatin- and GS-treated s.c. tumor mice was compared in three independent experiments. Gene expression in tumors from oxaliplatin- and GS-treated s.c. tumor mice had both similarities and differences. Expression profiles for 332 genes had >2-fold differences between oxaliplatin- and GS-treated s.c. tumor mice groups. "

Here is the article's link——enter link description here.

ADD REPLYlink written 7 months ago by BioLite20

Well, that method section is pretty poor. You basically have no idea what they did and therefore cannot reproduce things. That having said, you cannot even tell if their analysis is technically correct. That is all I can say based on the provided link.

ADD REPLYlink written 7 months ago by ATpoint15k

Yes, It can say that I made a very slowly and painful progress in Bioinformatics because I learning by myself completely.

Now, I am trying to practice analysis Agilent single channel microarray. AT, maybe you can give me some precious suggestions or another good practice article?

Thank you very much and best wishes to you!

ADD REPLYlink written 7 months ago by BioLite20

You should spend quality time reading the manuals of the established microarray analysis tools like limma and search around for blogs and tutorials. At least that is what I typically do.

ADD REPLYlink written 7 months ago by ATpoint15k

Yes, I'm doing like you said. Thanks again!

ADD REPLYlink written 7 months ago by BioLite20
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: 765 users visited in the last hour