Examining the effect of random damage and algorithm choice on the accuracy of DNA identification (Lepidoptera: Noctuidae)

Because certain groups of insects, such as the trifine Noctuidae, have many species, are difficult to identify if damaged, and have morphologically different lifestages, DNA represents an important tool for accurate species identification. DNA identifications are made by comparing a sequence fragment from an unknown individual to sequences of morphologically identified species. Comparisons are performed using a phylogenetic algorithm, typically Neighbor Joining (NJ), because of rapid completion time and unambiguous results. Other, more robust analyses such as Maximum Parsimony (MP) and Maximum Likelihood (ML), have not been widely tested for their efficacy in this process. The reliability of identifications from non-archival DNA is also suspect, because the DNA may have high amounts of ambiguity due to damage. This study simulates non-archival DNA sequences by creating “damaged ambiguous test sequences” (DATS) that have 5%, 10%, 20%, 30%, 40%, and 50% randomly missing base pairs. We generated these DAT’s from 64 noctuid sequences, generated from approximately five individuals per species, and compared them to a sequence database of these noctuids. Accurate identification to genus and species was recorded for each level of damage. Algorithm choice (NJ, MP, and ML) was also examined for each DAT. The implications of these results will be discussed in terms of the utility of DNA for species identification in real world settings.