Dealing with technical replicates

[deanie]

Thanks gringer. It does seem a little grey compared to other examples of technical and biological replicates I've seen. I expect the biological variation to be limited as it is a blood sample taken at once (i.e. fingerprick and three 10ul samples taken).

[gringer]

It is a type of technical replicate, but will have some degree of biological variation from the different samples. Usually "technical replicate" refers to an additional sequencing library prepared from the same extract.

In this case because there is a chance of biological variation, I would keep the samples separate and add the sample as an additional factor in any statistical models.

[deanie]

First question, can anyone confirm these are indeed technical replicates?

3 blood samples taken from the same individual, each processed separately and final libraries pooled for sequencing (paired-end). So have a pair of fastq files for each of the replicates.

Second question, what are your opinions on how to deal with the technical replicates during differential expression analysis?

I've heard of technical replicates sequenced over different lanes and then combining the fastqs. In this lab's experiment there is no lane variation due to the multiplexing. Is it still sensible to combine them in the case described above?

Or separately process the fastqs for each replicate (i.e. qc, trimming, counts) and then combine the read counts? Based on what I read most people sum the counts of the technical replicates, with a few posts suggesting averaging them.

TIA!