I found predORF from systemPipeR to predict the longest non-overlapping reading frames in a genome, however these are the longest ORFs per strand. I want to have the longest ORF considering both strands, for example:

AAAA**ATG**XXXXXX**TGA**AAAA     CDS = XXXXXX = 6 nt  <- pick this one
AAAATAGX**AGT**XXACT**GTA**A     CDS = XXACT  = 5 nt

Is there any package in R that can do this, I suppose it isn't that easy to code this myself since it requires quite some comparisons when both strands need to be considered.