The process of training the SOM neural network will adjust the weight gradually based on the input vectors. Results of training phase will create several clusters on the SOM output layer. The data objects with the proximity will be grouped into cluster that has the nearest distance from the input pattern to the neuron representing this cluster. According to the nature of the training algorithm of SOM neural network,the clusters that are near each other on the SOM output layer will contain data objects of a high similarity level ( proteins with similar structure) [5, 6].