How is it possible to recognize replicates?

Question

[Lorenzo]

Hi to all,

I'm new in analyzing RNA-seq data. I have to evaluate gene expression in different cells and tissues available on online datasets. To perform DEA I need replicates of course. How can I be sure that these below are technical replicates?

They derive from the same experiment, I mean same protocol, same day of library preparation, same day of sequencing and same sequencing technology but they are different runs. In your opinion are they replicates?

Thank you in advance for your time and suggestions!

Answer 1

[Carlo Yague]

These loos like these technical duplicates indeed. According to "library name" field under the ERR258415[0-1] run pages, both datasets are independant runs on the same sRNA library (called 683610_Occipital_Cortex).

But it might be best to contact the authors directly if you want to be 100% sure.

[Lorenzo]

Ok thank you, it is also my idea. Probably I will contact them. I have also another question. From a biological point of view does it make sense to compare tissues from different experiments (same protocol of RNA-extraction and technology of sequencing but different day of library preparation and of sequencing) or the risk of batch effect is too high?

[Carlo Yague]

There will be a batch effect for sure ! Sometimes it is small, sometimes big, but it is best to avoid it OR control for it (for instance, resequencing a bit of the first libraries along the newer samples).

[Lorenzo]

So the best option is to compare tissues or cells processed the same day to avoid batch effect (of course we cannot exclude it also in this case), right ? Instead to compare tissues or cells processed in different days can I proceed with an approach to estimate the batch effect? Actually I read different questions and posts about how to remove batch effect for example with PCA or edgeR

[Carlo Yague]

So the best option is to compare tissues or cells processed the same day to avoid batch effect ,

Yes

Instead to compare tissues or cells processed in different days can I proceed with an approach to estimate the batch effect?

Yes, this would be the second best option. But to estimate the batch effect, you need at least one sample sequenced across batches. Then it is possible to remove/control the batch effect. using edgeR, DESeq2, or something else.

[Lorenzo]

Thank you for your explanation, always very clear. What do you think about this bioconductor package http://bioconductor.org/packages/release/bioc/html/debrowser.html for less expert users? Have you ever heard about this?

[Carlo Yague]

I do not know that package, sorry

[Lorenzo]

Ok, thank you again for your clear answers!

[Carlo Yague]

if that was useful, you can accept my answer above :)