Clustering of fungal hexosaminidase enzymes based on free alignment method using MLP neural network

Abstract

Studies of biological evolution have generally focused on nucleotide or amino acid sequences of certain genes related to specific enzymes. Most phylogenetic tree constructions have been carried out using amino acid sequences and are used as a predictor to show evolutionary relationships. Phylogenetic analysis is usually performed based on multiple sequence alignment of a gene from different organisms including fungi. A number of programs have been introduced for gene clustering and phylogenetic analysis. For example, the most popular web-based program is Clustal Omega which is commonly used by biologists. When the number of uploaded sequences increases, this program not only works slowly but also the final constructed cladogram is confusing and incorrect from evolutionary point of view. In the present study, we used fungal hexosaminidases which are extracellular enzymes with a lot of applications in biotechnology but extremely varied and confusing in evolutionary terms. A standard taxonomy-based phylogenetic tree was constructed for 835 FH amino acid sequences retrieved from National Center for Biotechnology Information (NCBI) on March 16, 2015. Then a supervised multilayer perceptron (MLP) neural network was used to discriminate FH sequences. Based on relative frequency of amino acid in FH sequences, 41 neural networks were designed for seven levels from the phylum to family. Minimum accuracy of the neural network was equal to 99% at all seven discrimination levels. As a final step, an additional evaluation was performed on the designed model with 143 new released FH sequences extracted on July 1, 2015. The clustering results have shown a proper match with fungal taxonomy to show evolutionary relationships.