New algorithm of protein spot center detection in two dimensional electrophoresis gel images

Abstract

The most currently commonly-used approach of protein separation from the cell is two-dimensional electrophoresis (2DE). The proteins from the sample are separated according to their molecular mass and isoelectric point, facilitating afterwards to identify each of them individually. The analysis of the samples also includes protein registration and quantification. Many different approaches proposed for protein quantification use spot modeling. Mathematical models used for protein spots in 2DE gels differ with flexibility and simplicity of the model. The proposed in this paper spot modeling method uses radial basis network with anisotropic Gaussian basis function. The weakness of use of radial basis network for protein spot modeling is the estimation of number of units in the hidden layer of the network. The network is improved by the use of new spot detection in the hidden layer method.