A Genetic Programming Method for the Identification of Signal Peptides and Prediction of Their Cleavage Sites

A novel approach to signal peptide identification is presented. We use an evolutionary algorithm for automatic evolution of classification programs, so-called programmatic motifs. The variant of evolutionary algorithm used is called genetic programming where a population of solution candidates in the form of full computer programs is evolved, based on training examples consisting of signal peptide sequences. The method is compared with a previous work using artificial neural network (ANN) approaches. Some advantages compared to ANNs are noted. The programmatic motif can perform computational tasks beyond that of feedforward neural networks and has also other advantages such as readability. The best motif evolved was analyzed and shown to detect the h-region of the signal peptide. A powerful parallel computer cluster was used for the experiment.