Neural Computing

Besides the tracing of the pathway between genetic information and biological existence is not obvious and easy to see in a level of modeling aim, those relations are quite clear and confirmed in a qualitative  - theoretical – way. One may keep in mind that each bio-structure may be boiled down to genetic information, even though those are not simple enough for an easy-recipe implementation. The first bio-structure resulting from gene existence is called protein and its activity is modeled via transcription networks. Proteins have been found taking place in all biological processes, from DNA repair, protein synthesis to immune system. A quite important group of protein is called enzymes, they speed up reactions. A second quite important group is called transcription factors, they control gene activity.

            In the other hand, it has been studied intensively a group of mathematical models named Artificial Neural Networks, or simply, neural networks. They are models that have the brain as source for basic principles. In a short and not complete time line, the works have started in 1943 with McCulloch and Pitts proposing a general theory for learning machine; 1960, Rosenblatt and his collaborators studied the perceptron; 1961, Caianiello created a learning algorithm based on hebb’s learning rule; 1970, use of Evolutionary Algorithms by  John Holland; and 1985, recognition of the back-propagation by some group of scientists. In the current state of the art, those methods based on learning machine are applied intensively for modeling natural phenomena with high nonlinearity-level, nosy, but with some hidden law that are difficult or even impossible for the current mathematical theory captures. 

            Here is studied the application of neural networks for modeling gene expression. In a more precise way, gene expression is modeled via estimation, once the sharp value is not possible to be known, due to technological limitation and biological fluctuation.

From: Neural Computing.