Reference DNA databases for forensic species identification: Auditing algorithms

Reference databases are indispensable tools for species identification in forensic context. Among the most widely used public databases stands GenBank or BOLD, aggregating genetic data and associated metadata of millions of specimens. Acquiring and collecting such amount of information requires a concerted effort of the scientific community. Detection and correction of inaccuracies, such as erroneous/incomplete metadata or poor DNA sequences quality, as well as the species' sample representativeness and reliability, are of paramount importance. Here we implement an algorithm that can be used to automatically classify sampling reliability and access the quality of the compiled data allowing for the identification of possible sources of errors in a semi-automated auditing and annotation protocol. An annotation protocol was implemented to access the concordance between taxonomic assignment and automatic clustering groups resulting from molecular data. This was applied to a set of 943 barcodes of the Canidae family collected from BOLD database, due to its particular interest in forensics.