Sorry but this question is confusing on so many levels that it's barely understandable.
Predicting miRNAs.. the precursors? Or do you mean miRNA targets, and how exactly are you trying to predict them?
There's no such thing as non-coding or coding protein. Do you mean non-coding transcripts?
Blastx works on the assumtion that you try to map coding mRNA to a protein database. What would be the purpose of removing proteins in that context? What are you trying to achieve?