Neural network based phylogenetic analysis

Phylogeny is the primary tool used to understand the evolutionary relationship between various taxonomic groups. If someone finds an unknown bone fragment it is important to first identify which “class of animal” that fragment may relate to. The standard way would be to extract DNA, amplify and sequence a conserved region and construct a phylogenetic tree and then understand the Class to which it belongs. But this method involves various tasks such as multiple sequences alignment and constructing dendrogram by distance, parsimony or Bayesian methods, which requires considerable time and effort. In this research we implemented a probabilistic neural network to understand the possible class of animal of an unknown DNA sequence without using the phylogenetic tree approach. To achieve this target we used 90 Transferring sequence and 400 sequences of NADH dehydrogenase subunit I coding region of Mitochondrial DNA as data sets. The neural network was trained using the inputs created based on codon count extracted from the DNA sequences using tri gram method. The performance of the neural network based analysis was compared with phylogenetic analysis and the accuracy of the probabilistic and feed forward neural network approaches were also compared. Results revealed that the new approach performed better than the standard phylogenetic approach.