Consistency Of Biological Samples In De Novo Transcriptome Assemblies

Question

[Damian Kao]

I've heard that it is generally good practice to assemble transcriptomes from libraries of similar biological conditions (eg. assemble liver and heart samples separately rather than all together).

I guess the rationale is that having consistent composition of transcripts due to similar biological conditions let's the assembler distinguish between noise and real transcription? A transcript that is lowly expressed in one sample and normally expressed in another might make the assembler think it is just background noise? Or maybe different populations of isoforms in different conditions might confuse the assembler?

Is this generally a good rule to follow and what exactly is the reason behind it? Are there any publications regarding this?

Answer 1

[Biojl]

Actually in my experiencie, which to be honest is not much, I've seen that is a good idea to mix samples from the same individual (tissues, conditions) in order to build the reference transcriptome, getting more and longer (real) transcripts. What doesn't work is to mix different individuals, due to obvious reasons. I would be very interested in knowing the reasons behind your statement as well.