I used Trinity for de novo assembly with default parameters (k-mer 25). After that I used rnaSPAdes, which determine dynamically the best k-mer length for my data (kmer 73 and 49 in this case). I tried to compare both assemblies, but I didn't see many differences so I decided to continue my downstream analysis with Trinity, which is so much easy and user friendly.

My question is: now I know the better k-mer lenght for my data is 73 and 49, ¿May I use Trinity with this k-mer length? ¿May I merge rnaSPADes and Trinity assemblies and then use CD-HIT for redundancy?

Thank you in advance :)

The algorithm of Trinity is optimized for short k-mers. Brian Haas, an author of Trinity, doesn't recommend to increase the k-mer size (https://github.com/trinityrnaseq/trinityrnaseq/issues/1105)

You can try to merge two transcriptome assemblies. A pipeline called EvidentialGene can do this (http://arthropods.eugenes.org/EvidentialGene/trassembly.html).