Merging PE reads before de novo RNA-Seq assembly

[roussine]

Hello!
can someone please give an opinion: does that make sense to merge PE reads before a de novo RNA-seq assembly (say with Trinity)?
Merging tools are quite fast (like USEARCH www.drive5.com/usearch/manual/merge_pair.html), and the amount of raw data is reduced drastically. Can merged data be suitable for further assembly in a “single-ends” mode, or am I missing some theory behind it?.. Seems like it may save a lot of computations.

Thanks in advance.
Leo

[GenoMax]

For reference cross-posted to: https://www.biostars.org/p/295213/

[pmiguel]

I've never understood why people think that pair-merging prior to de novo assembly is a good idea. Seems like were that the case, assembly engines would just do pair-merging themselves prior to other assembly steps.

--
Phillip