DEseq2 tool on Galaxy software problems regarding normalization of featurecounts.

Question

[valid.universitadipavia.it]

Hi

I need help in the normalization steps of counts. I am using the DEseq2 tool for normalization but in the end, it gives me a lot of zero. I just want to know how I can normalize it again without getting so much zero numbers?

By the way, I used Galaxy software to normalize my feature counts. And software gives me automatic normalization but it seems not normalized.

Answer 1

[WouterDeCoster]

I just want to know how I can normalize it again without getting so much zero numbers?

You didn't really tell us about how many genes have zero expression, so you may have to be more specific about what you observe. It's not so weird to have genes with zero expression because not all genes are expressed in your tissue of interest.

[valid.universitadipavia.it]

Actually, I am doing a whole-genome analysis. So I just look at ~25000 genes. My sample counts did not appear or get 0 value in approximately half of the 25000 genes.

[WouterDeCoster]

I'd say that's totally normal.

[valid.universitadipavia.it]

I found also one interesting thing. When I normalize my sample counts by Using the DEseq2 toll I got lots of zero values but not values which are less than 0. But when I use the edger tool I got reverse results - less 0 value approximately even less than 100 but lots of value which are minus so less than 0. Do you have any idea about that?

[Devon Ryan]

How did you compute the normalized counts in edgeR? Likely there was an error there, since there's no equivalent to counts(dds, normalized=T) in edgeR.

[valid.universitadipavia.it]

You know I have used both tools on Galaxy software. And in the edgeR case, I have set up Min log2 fold change as 0 and of p-adjusted value 0.05. And normalization method I choice TMM. So probably what I think it chooses median and if the value is less than this median they are indicated as a negative value if it is the higher reverse case - positive.

[WouterDeCoster]

Now, which column are you looking at that has those zeros and negative values?

[valid.universitadipavia.it]

File looking is the following. I have an ID of the genes in the row and I have counts of my samples. Basically, I get lots of zero in columns overall. It seems that half of the genes are not expressed or expressed at a very low level. It is quite unlikely.

[WouterDeCoster]

It seems that half of the genes are not expressed or expressed at a very low level. It is quite unlikely.

I don't know which tissue you are working on, but mostly that is just very likely. You don't need kidney genes in blood, so these won't be expressed.

[valid.universitadipavia.it]

You know I am actually working on the cancer cell. I just wondering that have you ever used Galaxy software to do normalization by using the DEseq2 tool. Maybe I may some mistake in pre-run options.

[Devon Ryan]

You might want to ask about Galaxy on the Galaxy support site, it's not heavily used by the regulars here.