User: Nik Tuzov

gravatar for Nik Tuzov
Nik Tuzov70
Reputation:
70
Status:
Trusted
Location:
United States
Website:
https://www.linkedin.c...
Last seen:
1 week, 1 day ago
Joined:
3 years, 8 months ago
Email:
n*****@partek.com

Posts by Nik Tuzov

<prev • 55 results • page 1 of 6 • next >
0
votes
1
answer
134
views
1
answers
Comment: C: Estimation of non-trivial contrasts in MAST
... There is a definite difference between MAST in DESeq2 when designs with interactions are considered. In MAST, all of the regression coefficients are shrunken in the logistic part no matter what, whereas in DESeq2 the shrinkage of regression coefficients is disabled when interactions are present: ht ...
written 18 days ago by Nik Tuzov70
0
votes
1
answer
134
views
1
answers
Comment: C: Estimation of non-trivial contrasts in MAST
... Hello Andrew: What happens in contrast estimation if there is a quantitative covariate? E.g. I have the following toy data set: y <- c(0.0, 0.0, 0.01981825833, 0.40109357786, 0.34727231426, 0.03889038132, 0.0, 0.0, 0.11225699507, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.90902480961) ...
written 9 weeks ago by Nik Tuzov70
0
votes
1
answer
134
views
1
answers
Comment: C: Estimation of non-trivial contrasts in MAST
... Hello Andrew: You are probably familiar with [this paper of Soneson and Robinson][1]. Their overall conclusion is that the best single-cell-specific method was MAST, but it did as well as the best of bulk RNA methods. Do you think they got something wrong? [1]: https://www.nature.com/articles/nm ...
written 10 weeks ago by Nik Tuzov70
0
votes
1
answer
116
views
1
answers
Comment: C: Logistic regression in MAST - question
... Hello Andrew: Thanks for replying. If I don't want to create SingleCellAssay object, how can I put a number of transcripts in a data frame to do a similar analysis. Do I need to somehow add TranscriptID column to that simulated data frame? ...
written 10 weeks ago by Nik Tuzov70
2
votes
1
answer
116
views
1
answer
Logistic regression in MAST - question
... Hello: I am trying to reproduce an example given in MAST reference for zlm() where a data frame with 500 points is used. I assume this data set describes one single feature, hence no shrinkage is applied. I managed to reproduce the continuous part. The regression coefficients for the discrete pa ...
mast written 10 weeks ago by Nik Tuzov70 • updated 10 weeks ago by Andrew_McDavid190
0
votes
1
answer
134
views
1
answers
Comment: C: Estimation of non-trivial contrasts in MAST
... Hello Andrew: Thanks for replying. I also have a MAST specific question. The paper describes how shrinkage is done but it says little about why it was necessary in the first place. What kind of evidence did you get to justify the application of shrinkage? Regards, Nik ...
written 10 weeks ago by Nik Tuzov70
2
votes
1
answer
134
views
1
answer
Estimation of non-trivial contrasts in MAST
... Hello: Could you tell me if there is an easy way to estimate non-trivial contrasts in MAST? E.g. there are two factors, A and B, with 3 and 5 levels, the model is Y = A + B + A*B, and I want to test (A1, B2) vs (A2, B5). I assume the answer is no because a similar [question][1] was asked of DESeq2 ...
mast written 11 weeks ago by Nik Tuzov70 • updated 10 weeks ago by Andrew_McDavid190
0
votes
2
answers
1.2k
views
2
answers
Comment: C: Why limma, edgeR, and DESeq did not implement fold change moderation first?
... Is this issue present even when the design is orthogonal? ...
written 20 months ago by Nik Tuzov70
0
votes
2
answers
1.2k
views
2
answers
Comment: C: Why limma, edgeR, and DESeq did not implement fold change moderation first?
... The reason I asked the question was personal curiosity. Recently I started to review the older shrinkage methods (ridge, lasso, etc) that focus on shrinking the regression coefficients that describe the mean of response. At the time I had an impression that what popular DE packages did it "backwards ...
written 20 months ago by Nik Tuzov70
0
votes
2
answers
1.2k
views
2
answers
Comment: C: Why limma, edgeR, and DESeq did not implement fold change moderation first?
... I think in practice it's hard to separate these two. As far as I know, the user will need a list of genes with LFC estimates that are practically different from 0, and they have to be statistically significant (low p-value or high posterior odds ratio). ...
written 20 months ago by Nik Tuzov70

Latest awards to Nik Tuzov

Supporter 2.4 years ago, voted at least 25 times.
Popular Question 3.1 years ago, created a question with more than 1,000 views. For Normalization methodology in DESeq/DESeq2 vs edgeR

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 16.09
Traffic: 118 users visited in the last hour