Neural network based splice site prediction methods

Splice site is an important signal in genomic sequences that signifies the presence of coding region and plays an important role in gene structure prediction. In recent past, neural network based splice site prediction methods have shown remarkable success. The evaluation presented in this paper tries to identify most promising method based on neural network classification. Several measures are used to assess the predictive accuracy of these methods. The results imply that the predictive accuracy of the methods tested is lower than originally found which pinpoints that the methods are excessively reliant on the examples they learn from. Among all the method tested NetGene2 has revealed the best performance. However, the way extracted information is used in NetGene2 makes it a complex method for gene structure prediction.