This is certainly a topic for discussion with no definitive answers, but here are some suggestions:
-
De novo assembly
- Pros: useful if there is no reference genome for the species in question and/or if the annotation is of poor quality.
- Cons: the assembly if difficult and only the most abundant transcripts are likely to be fully assembled.
-
Mapping to the genome
- Pros: discovery of novel transcribed regions; de novo assembly of gene models for species with no annotation.
- Cons: limitations intrinstic to gapped mapping; analysis of the results in order to obtain transcript expression estimates slightly more complex.
-
Mapping to the transcriptome
- Pros: less complexity in the mapping (reads map contiguously); output easily interpretable; faster than mapping to the genome.
- Cons: only for annotated regions; multiple transcripts for the same gene (reads might map equally well to these because of shared sequence).