Is checking clustering of technical replicates necessary for RNAseq?

Question

[shl198]

Dear all,

I searched some threads, it seems like the standard recommendation for doing differential expression analysis is that: merge the fastq files of technical replicates together and then map to reference, count reads, do DE analysis.

My question is that will it be safer if we map, count and plot the cluster of samples, which make sure the technical replicates cluster together? After that we then merge the count data of technical replicates and do DE analysis.

[scottbrouilette]

Technical replicates only provide information about the technical variation - for example, how reproducible is your wet-lab methodology, or your sequencing platform. Assuming you are happy with these aspects there is no need for technical replicates - you then need true biological replicates to increase the power of your study. If you have technical replicates and they are reproducible then you can go ahead and merge the FASTQ files.

See the impact of various sources of variation in:

Lappalainen, T. et al. Transcriptome and genome sequencing uncovers functional variation in humans. Nature 501, 506-511 (2013).

And why increasing the number of replicates is sometimes better (in terms of power to detect differential expression) than increasing read depth:

Liu, Zhou, & White. RNA-seq differential expression studies: more sequence or more replication? Bioinformatics 30, 301-304 (2014).

Answer 1

[Antonio R. Franco]

I am very much interested in the responses, because I've been told that technical replicates were not required when doing RNA-Seq