Regarding your question, the best option would be map the MetaPhlAn2 database markers, rather than the consensus markers, against the references genomes of the species.
In this way, you will know the region on the genome where the marker comes from and if it is part of a CDS or not.
I hope I have been helpful.
Best,
Aitor