A novel basis function neural network

The paper presents a novel neural learning algorithm for analysing biological data with non-numerical features (amino acids). The algorithm is derived from the conventional radial basis function neural networks and is referred to as a bio-basis function neural network (BBFNN). The basic principle is to replace the radial basis functions by bio-basis functions, each of which is a biological amino acid sequence. The theoretical basis for this is that most biological amino acid sequences have preserved local motifs for specific biological functions. The application of this new neural learning algorithm to the analysis of biological amino acid sequence data sets shows two advantages. First, the computational cost has been significantly reduced. Second, the prediction accuracy has been improved.