The help text suggests the seeds must be proteins, but appears to have built in nucleotide seed sets and the next argument suggests the seed reads can be DNA:

$ oa buildgraph -h
...
  --seeds seeds         protein seeds; either a fasta file containing seeds
                        proteic sequences or internal set of seeds among
                        ['nucrRNAAHypogastrura', 'nucrRNAArabidopsis',
                        'protChloroArabidopsis', 'protMitoCapra',
                        'protMitoMachaon']
  --kup ORGASM:KUP      The word size used to identify the seed reads
                        [default: protein=4, DNA=12]

Reading python/orgasm/indexer/_orgasm.pyx it appears to attempt to auto-detect protein vs DNA seeds.

    cpdef dict lookForSeeds(self, dict sequences, int kup=-1, int mincov=1,object logger=None):

        cdef AhoCorasick patterns
        cdef dict matches
        cdef str k   
        cdef bint  nuc   

        nuc = all([isDNA(sequences[k]) for k in sequences])

        if nuc:
            if logger is not None:
                logger.info('Matching against nucleic probes')
            patterns = NucAhoCorasick()
            kup = 12 if kup < 0 else kup
        else:
            if logger is not None:
                logger.info('Matching against protein probes')
            patterns = ProtAhoCorasick()
            kup = 4 if kup < 0 else kup

        for k in sequences:
            patterns.addSequence(sequences[k],k,kup)


        patterns.finalize()

        #minmatch = 50 if nuc else 15
        minmatch = int(self.getReadSize() // (2 if nuc else 6))

        matches = patterns.scanIndex(self,minmatch,-1,mincov)

        return matches

Assuming the seeds can be DNA, clarify the --seeds help (and explain for example if an entire reference mitochondria could be used, or if the seeds should be fragments only, for example genes from a known mitochondria from a related species).