I'm looking for a UCSC Trascriptome file for hg19 and hg38 that can be used for psuedo and lightweight aligners like Salmon and Kallisto. I know that the Ensembl annotation can be found easily by looking for the cdna file, but I haven't had much luck in finding a UCSC version.
From what I read the UCSC version is a mrna.fa file available via their UCSC Browser Downloads, but I would like to double check.
It's mostly a question of whether you want to use refseq or not. If you do, then use refMrna.fa files. If you prefer a more frequently updated (and therefore less reproducible latter if you download the file again) dump from genbank then go for mrna.fa. I would suggest using refseq for something like this, it turns out that you get a little better results from Salmon (and presumably Kallisto, but we use Salmon much more frequently) if you use a transcriptome with only the most likely isoforms present.