Predicting Yeast Protein Complex Using Decision Tree

Bioinformatics is a new cross subject that using informatics technologies to process the biological data. Proteomics is an important aspect of it. The data mining methods in the computer science area are useful in analyzing large amount of data. We apply the decision tree, which is a widely used method of data mining, into bioinformatics area. Based on the collected data of the yeast protein, we use this method to predict whether a randomly selected yeast protein pair is of the same protein complex. We also construct the support vector machine system and the decision forest system that consists of many single decision trees to do the same prediction work. The results of these methods are compared and analyzed. Finally, we analyze the results from the biological point of view. We use Java programming language to implement the auto batch prediction application. As the prediction results show, our decision tree method gains a relatively higher accuracy. Our work provides a practical tool for the research of the yeast protein complex and the results we get show the probable way of the research in the future. Our work is a good investigation in applying the data mining methods into bioinformatics area.