The General Regression Neural Network to Classify Barcode and mini-barcode DNA

The paper presents the first stage of a study on characterization of species using very short DNA fragments from COI gene (barcode gene). Gene fragments, not complete sequence, are a common scenario working with degraded samples or in pres- ence of noise. The proposed technique is based on a novel prototype-based classification approach and a modified General Regression Neural Network (GRNN). The proposed system can use 200 bp (over 650 bp of the COI gene) even if constituted by blocks of 50 bp scattered in random positions of the gene sequence.