Identifying DNA Strands Using a Kernel of Classified Sequences

Automated DNA sequencing produces a large amount of raw DNA sequence data that then needs to be classified, organized, and annotated. One major application is the comparison of new DNA sequences with previously known classified sequences. In this paper we present a new approach to perform these comparisons. From a kernel of previously classified DNA sequences, we identify distinctive oligomers, or short DNA sequences, that are infrequent and thus highly unique within the kernel. We then search for the presence of these distinctive oligomers in the new unclassified DNA sequences. Their presence indicates a possible relation between a new DNA sequence and every previously classified DNA sequence that shares the distinctive oligomer. Ultimately, unclassified sequences are related to classified sequences with which they share the highest number of distinctive oligomers. We explain the details of our technique and show some experimental results in a kernel of immunoglobulin DNA sequences.