About BENN Backtranslation Tool
Abstract
Degeneracy in genetic codons is the main problem encountered in backtanslation of amino acid sequences to the proper gene sequences using soft-computing techniques. The proposed method is based on neural networks and neighbors of each amino acid in the protein sequence have been used as features to predict the correct codons (not single nucleotides) in the gene sequence, considering the results of the past works. In order to increase the number of features associated in the prediction and prevent both noise effect of the training patterns (same inputs for different responses) and overtraining in neural networks, Ensembles of neural networks were trained for each amino acid. Testing the accuracy by 2-fold cross validation method, BENN was able to correctly predict 82% of nucleotides of the test sequences.