I need to identify DEGs from RNA seq datasets provided in GEO. I have found the GEO2R tool to be quite useful, but it is not available for all datasets, for instance if the series matrix file is not available. Various other txt and excel files are given, varying from one submission to another. These include RPMK files, count files (in excel) etc. Can someone please suggest a method that I can use to do the DEG analysis without processing the SRA files from the very beginning. Is there some way I can use these counts and RPKM files to get a good idea of DEGs. Are the count data normalized?
Thanks in advance
Hi Shivaksh,
Geo2R is not dedicated to RNA-seq analysis, but to microarrays analysis. I would also strongly not recommend to use DESeq2 to RPKM normalized data, since DESeq2 uses its own normalization on raw counts data files from RNA-seq analysis. In the case of RPKM data, you should use tools from the Cufflinks / Cuffdiff software suite.
If a complete list of DEG have not been published, I don't see any easy way to get one without performing some kind of analysis. One way of getting either RPKM or raw read counts from many RNA-seq is to use the EMBL RNASeq-er (https://www.ebi.ac.uk/fg/rnaseq/api/) which re-analyzed more than 250,000 RNA-seq datasets with a standardized pipeline. Once you have the GEO reference, you should look for its SRA number and look for it in RNASeq-er. Good luck (again)!
Hi Shivaksh,
as long as the count data is available to you, analysis of differentially expressed genes is rather quick and there are several good packages out there that will facilitate your endeavour. Please be aware that RPKM (reads per kilobase per million reads) is a normalized unit whereas raw counts are not. Anyway, some tools might require you to use the raw counts as input rather than a normalized dataset.
A very quick approach for DEG-identification is DESeq2 ( https://bioconductor.org/packages/release/bioc/html/DESeq2.html), which is also characterized by an extensive documentation. You would just have to create a matrix summary of the experiment and try the basic settings of the package. This will already help you to get a feeling for differential expression and - in my experience - is a rather quick and straight-forward approach.
However, I am not quite sure what you mean with "...from the very beginning". Since you obtain count-data from GEO, you already skipped a considerable part of NGS experimentation ;)
Good luck!
Thanks so much Lukas for your reply.
I probably should have mentioned Im a complete novice to NGS data analysis and do not know R programming language. And since RNA seq analysis is not the primary objective of my project I was looking for an easier method. Is it possible to get DEGs from count using a software such as Galaxy? Or do you feel that DeSeq2 will be easy to pick up despite having no experience with R?
Also could you tell me if the count data provided on GEO is normailized?
Thanks again
Hi Shivaksh,
here's the point about DEG-analysis: It is an intricate statistical test you perform on a huge dataset. The nature of the data is not really settled - although it has become very common to model the problem with a negative binomial distribution and I think it is absolutely acceptable to do so - and there aren't any "shelf-solutions" for this analysis. I would argue that the Bioconductor pipeline and DESeq2 in particular are the closest you can get an "easy" method. I would refer you to this manual and recommend you to simply folllow it step-by-step:
http://www.bioconductor.org/packages//2.13/bioc/vignettes/DESeq2/inst/doc/DESeq2.pdf
Whether or not GEO data is normalized should be clear from the unit - if it has been normalized before distribution it should be a unit like RPKM.
R is a rather intuitive language, you will grasp it quickly.
Good luck!
Hi Shivaksh,
Geo2R is not dedicated to RNA-seq analysis, but to microarrays analysis. I would also strongly not recommend to use DESeq2 to RPKM normalized data, since DESeq2 uses its own normalization on raw counts data files from RNA-seq analysis. In the case of RPKM data, you should use tools from the Cufflinks / Cuffdiff software suite.
If a complete list of DEG have not been published, I don't see any easy way to get one without performing some kind of analysis. One way of getting either RPKM or raw read counts from many RNA-seq is to use the EMBL RNASeq-er (https://www.ebi.ac.uk/fg/rnaseq/api/) which re-analyzed more than 250,000 RNA-seq datasets with a standardized pipeline. Once you have the GEO reference, you should look for its SRA number and look for it in RNASeq-er. Good luck (again)!
- Badges
- Science method