[Simon Anders]

i don't understand why these packages are so unknown...

Very simple: Permutation-based tests cannot be used if you only have very few replicates, and this is usually the case. I am not aware of a single published dataset that compares samples from two different biological conditions and has more than four biological replicates per condition. You are unusually fortunate to have such a large data set.

Once you have many replicates, permutation tests are in fact the way to go.

[kopi-o]

they all work with permutations (non-parametric) and not with a statistical model (i.e. poisson, neg. binomial) and are therefore not sensitive to co-regulated genes.

How is the "permutation vs. model-based" distinction related to co-regulated genes?

[dietmar13]

i am no hard-core statistician,

but co-regulated genes affect (and disturb) distributional assumptions used in model-based approaches. the null-hypotheses generated by permutations are independent for every gene.

[Wolfgang Huber]

Dear Dietmar13,

let me respectfully disagree. Permutation tests, if applied properly, do respect the correlations between the genes and can model them as part of the null hypothesis.

Also, approaches that test gene-by-gene ("marginally", in the language of statisticians), including distribution- or model-based ones, are not directly affected by correlations between genes.

A more serious problem are correlations between samples, e.g.because of batch effects or poor experimental design.

Best wishes
Wolfgang

[wesserg]

RNA seq simulator output

Thank you all for interesting suggestions and comments.

@marcora: Indeed this dataset seems to be what I am looking for, but do you have any idea what kind of file format is it? Was trying several standard converters, but am not able to deal with it. Anyway - thank you very much for sharing the link;]

Problem:
I was also trying to generate myself such dataset using the RNA-seq Simulator from USeq bundle - but again - I have no idea where, the output format is specified. Anyone had previous experience with RNA-seq Simluator?

[Gordon Smyth]

I analysed my data with edgeR and I got 490 differentially expressed genes(300 upregulated and 190 down regulated). I am little bit confused. I am not sure the cut off value used to categorise up and down regulation. When I checked the fold changes, I could not find. If I want to do a pathway analysis, I am not sure which value I have to use ie fold change or p value. Can any body help me? I am not good is statistics.

Do you mean that you used decideTestsDGE() in edgeR to count the total number of differentially expressed genes? By default, that function uses FDR<0.05, as can be seen from the help page.

For remarks on performing a pathway analysis for RNA-Seq data, see:

https://mailman.stat.ethz.ch/pipermail/bioconductor/2012-April/044899.html

Gordon

[immpdaf]

Hi everyone,
I would be very grateful if anyone could give me some suggestions in our single-cell RNA seq data analysis part.

we have 2 groups of single cells (one normal single cells and one disease single cells), we performed single-cell RNA sequencing. Our library is made using SMART-SEQ2 protocol and it is single-end. We have around 4 million reads / single cell.

Now, using Differential gene Expression analysis, we are going to find significant genes which are upregulated or downregulated in disease cells group with regards to normal group.
So, which normalization technique could you recommend? Our bioinformatician uses TMM to normalize raw counts and he applies R package Monocle to perform DE.
He believes that if we use RPKM, we will get many false positive genes, since we are not comparing genes in one sample, but we are comparing different samples. Do you think it is right?

Many thanks in advance.