Moderator: Ryan C. Thompson
Ryan C. Thompson ♦ 7.3k
- Reputation:
- 7,260
- Status:
- Trusted
- Location:
- The Scripps Research Institute, La Jolla, CA
- Website:
- http://stackoverflow.c...
- Twitter:
- @DarwinAwdWinne
- Last seen:
- 3 months, 3 weeks ago
- Joined:
- 6 years, 7 months ago
- Email:
- r**@thompsonclan.org
Posts by Ryan C. Thompson
<prev
• 820 results •
page 1 of 82 •
next >
0
votes
1
answer
136
views
1
answers
... I'll assume you meant `T5.LD.FW` instead of `T5.SP`. If so, you are testing the T6 difference against the mean of the T5 and T4 differences. That's not quite the same as what you're after. In any case, I find that these "difference of differences" contrasts are difficult to interpret without the con ...
written 4 months ago by
Ryan C. Thompson ♦ 7.3k
0
votes
5
answers
241
views
5
answers
Answer:
A: Random Effect in DESeq2
... To my knowledge, none of the negative binomial GLM-based packages (e.g. edgeR & DESeq2) support models with random effects. Limma is the only package I know of that does support RNA-seq analysis with random effects. It supports a model with any number of fixed effects and one factor as a random ...
written 4 months ago by
Ryan C. Thompson ♦ 7.3k
0
votes
1
answer
136
views
1
answers
... What hypothesis are you attempting to test with this contrast? We can't tell you if your contrast is testing what you want to test unless you say what you want to test.
In any case, the contrast you have written appears to be subtracting two terms from one term, which usually yields a nonsense con ...
written 4 months ago by
Ryan C. Thompson ♦ 7.3k
2
votes
3
answers
258
views
3
answers
... Aaron's answer adequately covers the theoretical reasons that the normalizations used in edgeR and DESeq2 are appropriate for data with variations in globin content, so I will just add that empirically, I have actually *used* edgeR on such a data set. Specifically, we were testing a custom globin bl ...
written 4 months ago by
Ryan C. Thompson ♦ 7.3k
1
vote
2
answers
134
views
2
answers
... This isn't an answer to your main question, but for an in-depth discussion of how design matrices are constructed from factors, you should have a read through the [vignette for the codingMatrices package](https://cran.r-project.org/web/packages/codingMatrices/vignettes/codingMatrices.pdf). You gener ...
written 4 months ago by
Ryan C. Thompson ♦ 7.3k
1
vote
2
answers
181
views
2
answers
... You can read the code of the functions to find out exactly what they are doing. I did that a while ago but I don't recall the exact details. I just remember that the rule I came up with was "always run `eBayes` on every fit", and I haven't had any trouble since. ...
written 4 months ago by
Ryan C. Thompson ♦ 7.3k
0
votes
2
answers
181
views
2
answers
... It's only necessary to call `eBayes` if you need the statistics that it calculates. This will almost always be the case for the contrast fit (since you want to call `topTable` on it), but sometimes you also want it for the regular fit as well. It normally only takes a second or two to run it, so I j ...
written 4 months ago by
Ryan C. Thompson ♦ 7.3k
2
votes
2
answers
181
views
2
answers
... I recommend saving the result of `contrasts.fit` as a separate object, so that you still have access to the original `lmFit` result as well. They both have their uses, as you have noted. Run `eBayes` on both of them. My typical workflow looks something like this:
```
fit <- lmFit(mat_data, desig ...
written 4 months ago by
Ryan C. Thompson ♦ 7.3k
5
votes
2
answers
179
views
2
answers
... Looking at the code of duplicateCorrelation, the main time sink is the loop that runs mixedModel2Fit on each row of the input. So the time required should indeed be a linear function of the number of rows. However, the linear trend might break down if you are bumping up against hardware limits. The ...
written 4 months ago by
Ryan C. Thompson ♦ 7.3k
0
votes
2
answers
288
views
2
answers
... If you already have an alignment pipeline you're happy with, I wouldn't bother switching to another aligner. It's generally not going to make much of a difference. Just use featureCounts with your existing aligned BAM files. ...
written 4 months ago by
Ryan C. Thompson ♦ 7.3k
Latest awards to Ryan C. Thompson
Good Answer
6 months ago,
created an answer that was upvoted at least 5 times.
For A: Batch correction in DESeq2
Popular Question
7 months ago,
created a question with more than 1,000 views.
For summarizeOverlaps using GRanges or bed file as reads?
Appreciated
7 months ago,
created a post with more than 5 votes.
For A: Finding DE genes from RNA-seq data
Teacher
9 months ago,
created an answer with at least 3 up-votes.
For A: Voom transformation from counts and normalization to negative values
Scholar
9 months ago,
created an answer that has been accepted.
For A: Limma validity for only hundreds of genes/metabolites
Appreciated
9 months ago,
created a post with more than 5 votes.
For A: RNA-seq Normalisation - normalise all samples in experiment or only the ones use
Teacher
9 months ago,
created an answer with at least 3 up-votes.
For A: Voom transformation from counts and normalization to negative values
Teacher
10 months ago,
created an answer with at least 3 up-votes.
For A: What's the relationship between "design" and "contrasts"?
Scholar
13 months ago,
created an answer that has been accepted.
For A: Limma validity for only hundreds of genes/metabolites
Teacher
13 months ago,
created an answer with at least 3 up-votes.
For A: RNA-seq Normalisation - normalise all samples in experiment or only the ones use
Popular Question
13 months ago,
created a question with more than 1,000 views.
For edgeR Quasi-likelihood with tagwise dispersion?
Teacher
13 months ago,
created an answer with at least 3 up-votes.
For A: What's the relationship between "design" and "contrasts"?
Popular Question
13 months ago,
created a question with more than 1,000 views.
For summarizeOverlaps using GRanges or bed file as reads?
Scholar
13 months ago,
created an answer that has been accepted.
For A: Limma validity for only hundreds of genes/metabolites
Teacher
13 months ago,
created an answer with at least 3 up-votes.
For A: What's the relationship between "design" and "contrasts"?
Scholar
2.0 years ago,
created an answer that has been accepted.
For A: How does the function model.matrix (to define experimental design) really works
Scholar
2.0 years ago,
created an answer that has been accepted.
For A: Duplicate Correlation with technical replicates
Scholar
2.0 years ago,
created an answer that has been accepted.
For A: RNA-seq Normalisation - normalise all samples in experiment or only the ones use
Teacher
2.0 years ago,
created an answer with at least 3 up-votes.
For A: RNA-seq Normalisation - normalise all samples in experiment or only the ones use
Teacher
2.0 years ago,
created an answer with at least 3 up-votes.
For A: How does the function model.matrix (to define experimental design) really works
Teacher
2.0 years ago,
created an answer with at least 3 up-votes.
For A: What's the relationship between "design" and "contrasts"?
Scholar
2.0 years ago,
created an answer that has been accepted.
For A: What's the relationship between "design" and "contrasts"?
Appreciated
2.0 years ago,
created a post with more than 5 votes.
For A: RNA-seq Normalisation - normalise all samples in experiment or only the ones use
Popular Question
2.0 years ago,
created a question with more than 1,000 views.
For Rsubread::featureCounts extremely slow on some annotations
Use of this site constitutes acceptance of our User
Agreement
and Privacy
Policy.
Powered by Biostar
version 16.09
Traffic: 279 users visited in the last hour