SEQanswers

Go Back   SEQanswers > Search Forums


Showing results 1 to 25 of 91
Search took 0.04 seconds.
Search: Posts Made By: Gordon Smyth
Forum: Academic/Non-Profit Jobs 05-20-2015, 07:58 PM
Replies: 0
Views: 981
Posted By Gordon Smyth
Research Group Leader, Bioinformatics, Melbourne, Australia

Note: applications have now closed for this position

Laboratory Head – Bioinformatics

Walter and Eliza Hall Institute of Medical Research, Melbourne, Australia

Applications Close: Sunday 14...
Forum: Bioinformatics 04-22-2015, 05:41 PM
Replies: 5
Views: 19,341
Posted By Gordon Smyth
That is not quite the same as what edgeR's cpm()...

That is not quite the same as what edgeR's cpm() function does. It's somewhat better to offset the counts rather than the cpms.
Forum: Bioinformatics 04-22-2015, 12:01 AM
Replies: 5
Views: 19,341
Posted By Gordon Smyth
Which edgeR CPM do you mean in particular? Do you...

Which edgeR CPM do you mean in particular? Do you mean the column called logCPM in the topTags table? That is computed using the aveLogCPM() function.
Forum: Bioinformatics 03-30-2015, 01:57 PM
Replies: 4
Views: 3,386
Posted By Gordon Smyth
You cannot use RPKM with either edgeR or DESeq,...

You cannot use RPKM with either edgeR or DESeq, as the documentation for those packages will tell you.

All the packages (edgeR, DESeq, RUVseq) use read counts, which is what you should get when...
Forum: Bioinformatics 03-29-2015, 05:01 PM
Replies: 4
Views: 3,386
Posted By Gordon Smyth
That's a difficult problem. I'm suspect there...

That's a difficult problem. I'm suspect there might not be a good solution, but you might look at the RUVseq package and accompanying paper.



In my opinion, the expense of doing biological...
Forum: RNA Sequencing 03-25-2015, 04:22 PM
Replies: 4
Views: 2,047
Posted By Gordon Smyth
Computing the contrast is not the same thing so,...

Computing the contrast is not the same thing so, yes, you just could just compare the DEG lists. The same answer would apply in any of the packages (edgeR, DESEq2 etc).

If you have a large number...
Forum: RNA Sequencing 03-23-2015, 12:27 AM
Replies: 4
Views: 2,047
Posted By Gordon Smyth
I'm not quite sure what you are trying to do. Are...

I'm not quite sure what you are trying to do. Are you simply trying to test the interaction: (KO_CD - KO_SD) - (WT_CD - WT_SD)?

Have you looked at Section 3.3.1 of the edgeR User's Guide? The...
Forum: RNA Sequencing 03-21-2015, 11:28 PM
Replies: 5
Views: 2,188
Posted By Gordon Smyth
Your data sounds pretty crazy and it is quite...

Your data sounds pretty crazy and it is quite possible that TMM is not a strong enough normalization. You might try quantile, but it is hard to imagine any normalization that will robustly handle...
Forum: Bioinformatics 03-21-2015, 09:31 PM
Replies: 5
Views: 2,837
Posted By Gordon Smyth
Read the section on normalization in the edgeR...

Read the section on normalization in the edgeR User's Guide. Amongst other things, it explains why the pseudo counts should not be used as normalized counts.
Forum: Bioinformatics 03-16-2015, 03:33 PM
Replies: 5
Views: 2,837
Posted By Gordon Smyth
You can't output normalized counts from edgeR,...

You can't output normalized counts from edgeR, because the normalization factors are applied to the fitted model, not to the counts themselves. Of course you can output FPKM or counts-per-million,...
Forum: RNA Sequencing 03-08-2015, 11:32 PM
Replies: 2
Views: 2,965
Posted By Gordon Smyth
For a comparison that includes the t-test, see...

For a comparison that includes the t-test, see the voom paper at

http://genomebiology.com/2014/15/2/R29

As expected, the t-test does very poorly compared to the best of the RNA-seq specific...
Forum: Bioinformatics 01-29-2015, 08:47 PM
Replies: 154
Views: 58,377
Posted By Gordon Smyth
Hi Colin, We seem to be in complete...

Hi Colin,

We seem to be in complete agreement. Our group probably uses RPKM much the same as you do, for descriptive plots and expression summaries.

We don't output RPKM from featureCounts just...
Forum: Bioinformatics 01-28-2015, 12:32 PM
Replies: 1
Views: 824
Posted By Gordon Smyth
It is best to build up the condition-specific...

It is best to build up the condition-specific genes by making comparisons between the conditions.

Suppose you want genes that are specific to condition A. For a very stringent set, choose genes...
Forum: Bioinformatics 01-28-2015, 12:26 PM
Replies: 1
Views: 824
Posted By Gordon Smyth
Just put the numerical levels into a column of...

Just put the numerical levels into a column of the design matrix:

x <- numerical values
design <- model.matrix(~x)

Then do DE analysis for this coefficient.
Forum: Bioinformatics 01-28-2015, 12:11 PM
Replies: 154
Views: 58,377
Posted By Gordon Smyth
Counts are needed for a proper statistical...

Counts are needed for a proper statistical analyses, as is explained in the publications behind edgeR, voom and DESeq.

However, generating RPKM from featureCounts is easy, see:

...
Forum: Bioinformatics 01-28-2015, 11:59 AM
Replies: 5
Views: 2,811
Posted By Gordon Smyth
Type ?removeBatchEffect to read the documentation...

Type ?removeBatchEffect to read the documentation page. The documentation page explains that it is used to make unsupervised plots rather than for differential expression analyses.

This is the way...
Forum: Bioinformatics 01-27-2015, 07:24 PM
Replies: 1
Views: 1,161
Posted By Gordon Smyth
It depends what you mean by "adjusting for copy...

It depends what you mean by "adjusting for copy number". Your analysis is adjusting for baseline differences between the copy number groups, but it is also assuming that the logFC between Disease and...
Forum: Bioinformatics 01-27-2015, 07:14 PM
Replies: 5
Views: 2,811
Posted By Gordon Smyth
This is actually a very common type of RNA-seq...

This is actually a very common type of RNA-seq analysis where we combine two datasets. You can run voom and limma on the raw counts, as you would for any analysis. When you form the design matrix,...
Forum: Bioinformatics 01-26-2015, 05:10 PM
Replies: 2
Views: 1,136
Posted By Gordon Smyth
You can get a topTags table for all genes, then...

You can get a topTags table for all genes, then look at the ones you are interested in:

tab <- topTags(mydata, n=Inf)
i <- tab$GeneID %in% MyGenes
tab[i,]
Forum: Bioinformatics 01-26-2015, 04:55 PM
Replies: 4
Views: 1,200
Posted By Gordon Smyth
A simple and good way to handle genes of a priori...

A simple and good way to handle genes of a priori interest is by limiting the multiple testing adjustments. Normally you will run topTags(fit) on the whole dataset, where fit is output from either...
Forum: Bioinformatics 10-09-2014, 04:07 PM
Replies: 7
Views: 3,223
Posted By Gordon Smyth
cpm is a very simple quantity: count /...

cpm is a very simple quantity:

count / normalized lib size * 1e6

It doesn't depend on dispersion.
Forum: Bioinformatics 10-09-2014, 02:20 PM
Replies: 7
Views: 3,223
Posted By Gordon Smyth
Note that to access the pseudo counts, one uses ...

Note that to access the pseudo counts, one uses

dge$pseudo.counts

not

equalizeLibSizes(dge)$pseudo.counts

The latter code will (in Bioc 2.14) compute pseudo counts using dispersion=0....
Forum: Bioinformatics 09-23-2014, 04:29 PM
Replies: 7
Views: 3,223
Posted By Gordon Smyth
Nacho, thank you for sending me your data...

Nacho, thank you for sending me your data offline. To my surprise, this turns out to be neither impossible nor a bug.

First, let me point out that this would not have occured had you used edgeR in...
Forum: RNA Sequencing 09-22-2014, 07:02 PM
Replies: 2
Views: 1,291
Posted By Gordon Smyth
No it doesn't. Using 1,-1 instead of -1,1...

No it doesn't.

Using 1,-1 instead of -1,1 gives exactly the same genes just with logFC of opposite sign.
Forum: Bioinformatics 09-22-2014, 12:01 AM
Replies: 7
Views: 3,223
Posted By Gordon Smyth
nachocab, the results that you report sound...

nachocab, the results that you report sound impossible. It isn't possible for a real count of over 5000 to be reduced to a pseudo count of 0. Either you've made a mistake somewhere or there's a bug...
Showing results 1 to 25 of 91

 


All times are GMT -8. The time now is 03:16 AM.


Powered by vBulletin® Version 3.8.9
Copyright ©2000 - 2019, vBulletin Solutions, Inc.
Single Sign On provided by vBSSO