Backpropagation learning on ribosomal binding sites in DNA sequences using preprocessed features

Several studies have explored how neural networks can be used to find genes within regions of previously uncharacterized deoxyribonucleic acid (DNA). This paper describes the creation of a neural network training set for determining which part of a DNA strand codes for an important genetic feature called a ribosomal binding site (RBS). Based on previous research on detecting other genetic features, this data set contains preprocessed features that reflect biologically meaningful patterns in the raw base pair [ACTG]* language. We also describe preliminary empirical results indicating neural network performance that is superior to all other automated methods for detecting RBS´s