How much importance is RNAseq experimental bias?

Question

[John]

Hi there,

We did RNA sequence for 6 different treatments. Samples are from specific region of brain. But he took 5 mouse brains from one treatment and 7 mouse brains from other treatment (Reason: Ribotag IP gives 1 nanogram of RNA for 5-8 brains). Will that create big batch effect, Even if I normalize the data using normalizeBetweenArrays {limma} or something like this??

To be precise: Ribotag IP performed by pulling down ribosomes which expresses particular gene, then sequenced by SMART-Seq.

Have anyone came across this situation?

thanks

[ATpoint]

So you pooled 5 or 7 brains, pooled the RNA respectively and then prepared the libraries? Do you have replicates or is this then n=1 for every condition?

[John]

yes, I have two replicates for very treatment.

Answer 1

[Devon Ryan]

The difference in the number of pooled brains isn't a particular problem, though hopefully those were pooled from litter mates and then litters used as biological samples. What you should watch out for is if some brains took longer to extract and process than others, since that will certainly lead to batch effects due to RNA degradation.

[John]

okay, but the amount of RNA from 5 brains will be different from 6 brains right?

When I look at the read counts/TPM matrix, column sums are higher in the sample pool of brains than 5 brains.

[ATpoint]

It will but as you input a fixed amount of RNA to the library prep it should not matter too much. For the uneven read number, this is normal as sequencing never produces 100% even read counts. This is not necessarily an issue based on RNA amount. This is where normalization comes into place.