There are several ways you can try to predict the function of a sequence. A much-used method is to search for homologs with a known function, either in other species (orthologs), or in the same species (paralogs). Do you think Blast is a good tool for this? What kind of hits would you look for when searching for homologs? Think about significance (E-value: the probability of matching a random sequence in the database by chance) and annotation (description, references). blastx searches protein database using a translated nucleotide query (in all possible 6 reading frames). In other words, blastx gives you the closest protein sequence your nucleotide sequence will translate into (obviously if it is in the database).