Ideal settings for blasting a protein/cdna against a genome?

Question

[cmdcolin]

Are there any ideal settings for BLAST to align a protein or cdna against a genome? Or is it perhaps even misleading to rely on BLAST this way and it's better to use something like exonerate (protein2genome/est2genome)?

[h.mon]

Ideal settings are context-dependent: are the proteins/cDNA from the same species as the reference genome, or from a somewhat distant species? What is your purpose, finding close homologs or even distant homologs?

edit: anyway, I would use Spaln or Exonerate for proteins, and GMAP or Spaln or Exonerate for cDNA.

[cmdcolin]

It would be same species or fairly closely related (e.g. getting a danio gene, blasting against fish of interest). I don't necessarily have exact use cases but there were biologists wanting to find the location of their gene of interest in a poorly annotated genome, and they had cdna sequences from either experiment or a closely related species, and they wanted to blast against it and then look at it in the genome browser. I just noticed when trying to help that blasting a cDNA against the genome would get lots of similar paralog hits but might not have been obtaining the top matches as the right gene (or sometimes it would not even be in list) using default BLASTN of a cdna to the genome

[h.mon]

If that is the case, maybe the best approach is to iterate over blast algorithms: start with megablast; if nothing found, try dc-megablast; if nothing found, try blastn - of course, this is only feasible if done by hand with a few genes.

But the problem may be with the assembly, it may be too fragmented and the gene is not present or is fragmented between several contigs.