Something better than GSEA?

Question

[Matt.]

Hi,

I am new here and I have a question for you guys: which is the best tool to predict if a specific list of genes is up- or downregulated in your RNA-SEQ data? I am trying GSEA but I was wondering if there are better tools available.

Thanks a lot in advance for your help.

Matteo

[Amitm]

hi,

GSEA is for interpretation of gene lists in terms of enriched biological processes. For finding the gene list (up-/ down-reg) at the first place, use BioC packages like DESeq. Also important is whether you have replicates or not.

Answer 1

[EagleEye]

GSEA is an functional enrichment tool. Use tools for differential expression analysis like edgeR, DEseq, DEGseq, NOISeq etc.,

For choosing the best method, have a look at these articles where they compared different differential expression analysis packages.

1. A comprehensive evaluation of normalization methods for Illumina high-throughput RNA sequencing data analysis

2. Comprehensive evaluation of differential gene expression analysis methods for RNA-seq data

3. Evaluation of methods for differential expression analysis on multi-group RNA-seq count data

[John]

I'm waiting for the day when someone's software-comparison-paper-conclusion ends in "The data shows there was one obvious winner, prompting us to ask why our study was even needed in the first place."

[informatics bot]

Even if one software is inherently better than all the others, half of the bioinformatics community will continue to use the lesser ones.

[ivivek_ngs]

+1 for the screenshots at John. Difficult call but someday may be.

Answer 2

[Charles Warden]

GAGE and goseq are two Bioconductor packages designed specifically for RNA-Seq data:

http://bioconductor.org/packages/release/bioc/html/gage.html

https://bioconductor.org/packages/release/bioc/html/goseq.html

The GAGE manual even includes code to facilitate downstream analysis with DESeq (or edgeR or limma). goseq includes code that works with results from edgeR.

Functional enrichment tools that aren't specifically designed for RNA-Seq will probably also typically work OK, but I gather that is not what you were asking about.

Answer 3

[Matt.]

Any help, guys?

[ivivek_ngs]

You have to be more specific about what kind of experimental questions you are about to address. I believe you have 2 groups and triplicates then 4 samples means total of 12 for each group. So the DE analysis should be Cond1 vs Cond2 right. Your experimental design is a bit confusing still I would give some suggestions which can be followed. If you have 2 groups , lets say normal vs tumor or untreated vs treated and you run DESeq2 to get list of transcripts or genes that are deferentially expressed. In that case you will have both up and down genes for your experimental condition. Now what to do with this list? You can do a lot of stuffs, steps:

1. IPA analysis having both up and down list of genes along with FC values to exactly have pathway enrichment plots where you can see which pathway is activated or repressed and how your genes are enriched in that pathway.
2. Pathway analysis with Reactome, Ipath2
3. GO analysis with PANTHER,Amigo2,Enrichr
4. Upstream transcriptor factor analysis as well to find a master regulator. It is there in Enrichr and IPA.
5. Gene set enrichment analysis with GSEA
6. You can you the list of genes that are DEGs and feed in cbioportal if they are from human data and try to see the level of dysregulation these genes have across all cancers, or specific cancers
7. You can also use OASIS tumor tool to generate PAN-Cancer report for gene list or select any specific pathway of interest to see how these genes behave in that pathway

[zizigolu]

sorry, I searched in literature but I got confused because I saw they used whole of expression datasets as input for GSEA but here you noticed we should used only extracted differential expressed genes from our dataset as GSEA input may you please mention what is correct , DE genes or whole datasets as input?

thank you

[ivivek_ngs]

There are different way to do a GSEA analysis, it really depends upon your hypothesis that you are asking. Take a look at this link and in this pdf . You can use gene sets as well. It entirely depends upon what you want to see. You have to read the tutorial carefully.

[zizigolu]

I just finished differential expression analysis with cuffdiff. as i read in biostars posts, in gene_exp.diff folder of cuffdiff output we can see up or down regulated genes easily. in column 9 for example is sample 1 , in column 10 is sample 2 and if in column 11 we have negative value it means the gene expressed more in sample 2 and if the value is positive it means the gene expressed more in sample1 and this is proved if we have "yes" for this genes in column 14. then you able to select genes with positive or negative values in separate group. you can consider my latest posts because the codes for extraction the genes are there.

Answer 4

[Matt.]

I used DEseq, but what I need is: I created my personal list of genes of interest and I was wondering if there is a way to verify if these genes are up- or down regulated not as single but as group. (I have 4 samples each in triplicates). Thanks

Answer 5

[Matt.]

Maybe can IPA do this kind of analysis?