This seems error prone. Any mutation will damage integrity. Translating to amino acid sequences would be more reliable, since they use redundancy to protect against the more likely mutations. You could order the amino's by that redundancy - the ordering then also makes it easy to set a desired level of integrity protection by choosing how many amino's to use.
What you really want to use is an error-correcting code [1]. These have been applied to information storage in DNA [2].