Detection Of Gene Deletion Based On Artificial Neural Networks

With the rapid development of science and technology, the fusion degree among different subjects becomes higher and higher. Consequently, some new subjects are coming out. With the powerful operation ability and storage capability, computer technology has almost penetrated into all scientific fields. Bioinformatics is a new subject which is composed of life science and computer science and with the help of computer technology to accomplish the complex biological information acquisition, data processing, data storage and analysis. Because of the rapid development, bioinformatics has become one of the significant research fields in the21st century.Since being proposed by three American scientists in the year of1956, artificial intelligence as one of important branch of computer science has experienced several decades of development. Artificial intelligence has become one of the21st century cutting-edge technologies, which is widely used in intelligent recognition and robotics. In the life science field, genetic engineering also is known as one of the21st century cutting-edge technologies. Because genetic engineering has large data, complex research content, the research of genetic variation is an important content of genetic engineering and the close relationship with major serious diseases, so it is important to study genetic engineering. There are four kinds of genetic variation:deletion, insertion, inversion and duplication. It is significant to combine artificial intelligence with bioinformatics to detect whether there is gene deletion or not.This paper combines classification algorithms of artificial intelligence with detection of gene deletion of gene engineering. The specific tasks of this paper are as follows.(1) Analysis and find out the three shortcomings of Adaboost algorithm:the properties of experimental samples are not obvious, the uneven sampling method in the training process and the rapid growth speed of the misclassification sample’s weight. In view of the above shortcomings of Adaboost algorithm, this paper proposes SWA-Adaboost algorithm.(2) Using the8datasets of Machine Learning database of University of California.US, the article runs experiments to compare the original Adaboost with SWA-Adaboost. Although the classification accuracy of the modified algorithm is not as good as the original algorithm based on some datasets, but is better than the original one based on the most datasets. The experimental results show that SWA-Adaboost has better generalization performance than the original Adaboost and the average decrease of generalization error is about10%.(3) Combining machine learning with bioinformatics. This paper uses three algorithms of machine learning:Adaboost, SWA-Adaboost, and SVM. Extracting characteristics with biological significance for classification algorithms in machine learning, then using the three classification algorithms mentioned above to classify. Finally, get the ultimate generalization error of the Boosting series algorithms according to the generalization error of Adaboost and SWA-Adaboost, and the ultimate generalization error of SVM. The experiment result shows that the three algorithms of machine learning can be used in detection of gene deletion, and can achieve better detection effect.