Without the use of either a closely related genome, or multiple more distant relatives with similar gene sets, for reference genomes in annotation, it is very difficult to predict which ORFs are: true genes, gene fragments, or random small ORFs.