smithlabcode / riborex Goto Github PK

the CHANGELOG.md file should be up-to-date with the latest version of riborex (2.4.0) and should reflect the changes to all versions made from 2017 to 2019

By default, R set the levels of factor in alphabetical order, so no matter the condition you provide is

cond <- c("normal", "normal", "KO", "KO")

or

cond <- c("KO", "KO", "normal", "normal")

The base level will be always "KO". The current manual might be confusing with this issue, and the relevel function in R still causes warning because of the way the model.matrix is used in Riborex. We are working on a new version to make the base level more clear and intuitive.

Hi!
Thanks for the package, it is very useful and clear!

I was wondering if it is possible to take into account that the rna and ribo samples are paired. I am asking this becuase I have three replicates and one replictae is a bit different than the others in both, RNA and RPF sequencing, so I think that providing the information that the RNA and RPF samples are paired will improve the results.

I tried with the multi-factor function but it says the matrix is not full rank. This is the conditions matrix for both datasets (rna and ribo):
treat samp
cont1 control 1
cont2 control 2
cont3 control 3
treat1 treatment 4
treat2 treatment 5
treat3 treatment 6

If I number the treatment from 1 to 3 then I have a matrix of full range but the results do not have sense, as the control and the treatment samples are not really paired.

I would like to know if my experimental design is possible.
Thanks for your help,
Neus

Hi there,

Extremely thankful for your very easy to use Riborex package. Results are really interesting. I was wondering if I'm able to retrieve the normalized counts from any of the engines? Using DESeq2 at present but would also like to try edgeR and Voom. Can you provide this in the package?

Also in the vignette you have "control" ordered before "treatment" although in the results output it is "treated vs control". This is not what I get when running the code. Does "treatment" have to be before "control" in order to get a "treated vs control" output in results for riborex?

Kindest regards and thanks for your great work,
Marina

I'm trying to install riborex following the README and got this:

Warning in install.packages :
  package ‘riborex-1.2.3.tar.gz’ is not available (for R version 3.3.1)

Which version of R does riborex support?
Thanks!

riborex releases should be synchronized on github, CRAN and conda.

Currently the only data inlcuded with the repo is in the RData format, and it would be useful to have data that is human-readable so users can run an example and also compare their input data format with one known to work.

Thank you for your great package to analysis differential translation events on ribo-seq data. I am using your package to analysis my own data,but my rnaseq and riboseq data do not have replicates,how can i use this package to find differential translation genes ? I am looking forward to your reply!

Hello developer!
I'm trying your riborex for some ribo-seq data/rnaseq data that I have.
In my typical rnaseq workflow, as I work with stringTie-prepDe.py, I would make a DeseqDataSet object with DESeqDataSetFromMatrix. Then, I would apply, if necessary, a filter such as:
#dds <- dds [rowSums(counts(dds))>1, ]
or
#keep <- rowSums(counts(dds) >= 5) >= 3
#dds <- dds[keep,]
if I wish to remove the genes that have all 0's or so, else, I would
#dds <- dds [rowVars(counts(dds))>1, ]
to keep those genes that have (>1) variability.

I would then, revel so that the $condition in the dds object is correctly chosen.

Just after doing all of this, I would use the DESeq(dds) function.

I was wondering, if you took into consideration these or any other type of "data filtering" techniques. If so, how can I apply them? Or else, I would like to double check, if I filter over the DESeqDataSet object and then feed the filtered data with counts(dds) or assay(dds)

as appropiate for:
RNACntTable or
RiboCntTable

would that be ok?

Many thanks for your support,

MECC

Dear all,

I have been wondering about your reasoning to hard-code the lower boundary of the "minMeanCount" parameter to 1, as it prevents me from running a consistent analysis (= using exactly the same transcriptome) when testing the effect of multiple different knock-outs on my cells (total of 4 KO vs. wild type pairs).
It is obviously possible to compare the results of each individual analysis, however, since the background set of genes is never fully the same, combining all different conditions into one overview table via cbind() is for instance not possible and extra effort is required to match the different result lists accordingly.
Since my analysis are based on DESeq2 and the issue of testing low count genes using DESeq2 has been discussed previously (see for instance bioconductor mailing list), I do not see any reason as to why "minMeanCount = 0" should not be allowed.

Kind regards