Feedback requested for pooling biological samples for NGS

Question

[kumars.sv]

Hi all,

I have a question regarding NGS experiment set-up and sample pooling. We are currently planning WES and RNAseq of human samples between normal and tumor groups. We will also confirm the same in cell lines. Experimental setup suggested by PI is to pool 3 replicates in to one sample per group and do NGS i.e extract DNA/RNA from 3 samples (N=3 per each group, but after pooling 1 per group), pool them and do a pooled single sample NGS per group. However, I think this approach is incorrect. In my observation, it is like pool three samples and take one measurement. However, PI says it is correct as each single sample is a pooled sample of 3. It should not be a problem as there are several manuscript to support this kind. Any insights or link with detailed explanation (if it is correct or not) would be grateful. I wish I could have consulted with statisticians/ biostatisticians around. However could not find one familiar with such HTS experiments.

Useful manuscript is: https://bmcgenomics.biomedcentral.com/articles/10.1186/s12864-015-1767-y. However, authors use 2/3 pools per group with each pool containing 2-3 samples. In my case, there is only one pool per group and each pool with 3 samples.

PS: Most of the other PIs do the same and published the results.

Answer 1

[WouterDeCoster]

As usual, just because others do it and it has been published, doesn't make it right. Indeed, it is not the best idea to make pools of samples - they're not replicates anymore, you'll just get one signal per pool and can't figure out if this is caused by all or just one sample in the pool. In a way you do neutralize the most extreme outliers, generating a more 'average' signal, but you lose all information of biological variability. Can't you just use adapter barcodes in your experiments and keep the samples split?

[kumars.sv]

We send the samples to commercial providers. 3 samples without pooling are charged for 3 samples. Same problem as biotech deepti from A: question on sequencing strategies and batch effects and kumar issue from A: RNA Sequencing Pooling design and statistic analysis. Commercial vendors promise statistical significance and differential expression from one sample per group and PIs pool 3 samples to one sample per group and then do sequencing. Thanks for response.

[WouterDeCoster]

Their promise of differential expression is wrong, however, their charging does make some sense because they need to create a separate library per sample.