Most tools requrie a FASTA file containing the DNA or AA sequences as input. To start with, we are going to create a FASTA file containing the DNA sequence of the regions annotated in the GFF3 ("chr1.gff") file given a FASTA file ("chr1.fna"). One could expect that there are many tools that take a gff file and a fasta file and give you the sequences, but I haven't found any script that worked, not claiming there is none. The reason is maybe, that it is quite a simple task. One could also use Bioperl and work with the BIO::TOOLS::GFF module, but in this example I use R/Bioconductor which is maybe a bit slower but much less code is required. If you don't like this example, I have others. You need the packages rtracklayer and Biostrings installed.
Annotation software like Artemis might also be albe to do this.
Extracting regions into a FASTA file in R
library(rtracklayer)
library(Biostrings)
gff = import.gff3("chr1.gff")
fasta = read.DNAStringSet("chr1.fna")
myview = Views(fasta[[1]], start=start(gff), end=end(gff),
names= make.names(gff$ID, unique=TRUE))
myview[strand(gff)=="-"] <- reverseComplement(myview[strand(gff)=="-"])
write.XStringSet(x=DNAStringSet(myview), file="out.fas")
One important point to mention is that you need good names in your fasta file, so better have your ID field contain something meaningful, or use more fields from the GFF file.
Translating sequences
You can use Transeq from Emboss tools and translate in all 6 reading frames. For example, to search PFAM for functional domains in your translated sequences. Some tools (like Blastx) will do the translation, then you won't need to translate. For use with Blast (and other software that handles the translation) it is better to let that software handle this step.
Why do I recommend to translate in all frames? Because your annotation of gene starts and exon/intron boundaries might not be perfect yet, getting into the wrong reading frame. Therefore it is safer to check all 3 (or 6, as it is default in transeq, including reverse strand, that won't hurt).
Hope this helps.
Are the DNA sequences contained in the gff file?
As Michael says, you've told us everything except the most important thing: what format your "gene predictions" are in. If you have some form of nucleotide sequences, translation to protein could not be easier, but we need to know precisely what you have.