User: maltethodberg

gravatar for maltethodberg
Reputation:
20
Status:
New User
Location:
UCPH
Last seen:
1 day, 14 hours ago
Joined:
11 months, 1 week ago
Email:
m************@gmail.com

Posts by maltethodberg

<prev • 27 results • page 1 of 3 • next >
0
votes
1
answer
76
views
1
answer
Replacing values in an RleList
... Given an RleList (i.e. from either calling coverage() or import.bw()) what is the best way to replace certain values? I am interested in something this: x #RleList x[x < 0] <- 0 I'm doing this many times, and compared to how quick most other operations on RleLists are, this step becomes th ...
rlelist ifelse written 16 days ago by maltethodberg20 • updated 11 days ago by theobroma2210
0
votes
3
answers
232
views
3
answers
Comment: C: DE for genes with very low counts using limma.
... Played around a bit with the different thresholds using limma-trend, to see how it it affects the prior fit for lowly expressed genes. Made the following observations: 1) Filtering out genes with a substantial amount of zero counts across samples, removes most of the prior dip at lowly expressed gen ...
written 12 weeks ago by maltethodberg20
0
votes
3
answers
232
views
3
answers
Comment: C: DE for genes with very low counts using limma.
... That was more or less exactly what I was doing, ranking genes with aveLogCPM and then filtering out the most lowly expressed genes until the prior curve from plotSA looked decent. I will try playing around with different filtering criteria. With limma-trend my reasoning was that you could potentiall ...
written 12 weeks ago by maltethodberg20
0
votes
3
answers
232
views
3
answers
Comment: C: DE for genes with very low counts using limma.
... What specific approach of edgeR would you recommend using in this case, the standard or QL pipeline? ...
written 12 weeks ago by maltethodberg20
0
votes
3
answers
232
views
3
answers
Comment: C: DE for genes with very low counts using limma.
... Each treatment group is small (duplicates or triplicates). Several treatment groups are in larger blocks to control for various batch effects. ...
written 12 weeks ago by maltethodberg20
9
votes
3
answers
232
views
3
answers
DE for genes with very low counts using limma.
... I am analyzing a very large RNA-Seq experiment (100s of samples). I intend to use limma for the analysis, mainly due to the speed of limma compared to edgeR or DESeq2. I specifically want to keep a large number of very lowly expressed genes in the analysis. When I apply the voom, the fitted mean-va ...
voom limma limma-trend written 12 weeks ago by maltethodberg20 • updated 12 weeks ago by Gordon Smyth29k
0
votes
1
answers
121
views
1
answers
Comment: C: Design matrix and contrast for RNA knockdown experiment
... In that case, what would an appropriate FDR correction procedure be? In my understanding, most methods assume a  uniform distribution of p-values towards 1.0, which is not the case here. ...
written 3 months ago by maltethodberg20
0
votes
1
answers
121
views
1
answers
Comment: C: Design matrix and contrast for RNA knockdown experiment
... Just implemented the intersection union test. In some cases this gives rise to a p-value distribution heavily skewed towards 1.0 (similar to the scenario D here: http://varianceexplained.org/statistics/interpreting-pvalue-histogram/). Any idea of why this happens? ...
written 3 months ago by maltethodberg20
0
votes
1
answers
121
views
1
answers
Comment: C: Design matrix and contrast for RNA knockdown experiment
... Thank you for your detailed reply. I like your suggestion with the intersection union test, since that still produces a p-value for every gene. To test whether the mean effect across all oligos for a single gene is significant, would the following contrast for the Gene1 be correct?: c(0, 1/3, 1/3, 1 ...
written 3 months ago by maltethodberg20
4
votes
1
answer
121
views
1
answer
Design matrix and contrast for RNA knockdown experiment
... I intend to use voom+limma to analyse an RNA knockdown experiment that essentially looks like this: design <- data.frame(gene=c(rep("control", 2), rep("gene1", 6), rep("gene2", 6), rep("gene3", 6)), oligo=c(rep("control", 2), paste("oligo", c("a", "a", "b", "b", "c", "c", ...
contrast matrix limma design matrix written 3 months ago by maltethodberg20 • updated 3 months ago by Aaron Lun13k

Latest awards to maltethodberg

No awards yet. Soon to come :-)

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.2.0
Traffic: 127 users visited in the last hour