人類基因組計劃的開展隨之產生了巨量的基因組信息，區分DNA 序列上的外顯子和內含子成為基因工程中對基因進行識別和鑒定關鍵環節之一[1]。如何建立良好的系統模型將基因組數據進行有效地存儲、分析和挖掘，仍是難題。本文著重研究將多層前饋神經網絡應用于基因序列的預測分析中，成功從基因序列上識別出剪接位點，進而區分內含子和外顯子邊界。使用MATLAB 神經網絡工具箱和圖形用戶界面開發技術，對UCI機器學習數據庫中的基因數據集采用二進制數字編碼，完成樣本選取；創建優化算法的BP神經網絡和GRNN 神經網絡并加以訓練、仿真和測試。
關鍵詞 神經網絡；基因工程；內含子；外顯子
Abstract With the development of HGP, lots of genome information is acquired. To divide
the exorns and introns in DNA sequence is one of the key points of genetic identification in
genetic engineering. How to make a system model to storage，analyze and digging the genome data is still very difficult. This paper concentrates on how to use multilayer feedforward NN to getting primate-junction from genome sequence, and then, get the boundaries between exorns and introns. By using neural network tool box of MATLAB and GUI technology, to encode the data set from UCI Machine Learning Database with binary digit code, to gain examples.The designed BPNN model and GRNN model are used to train, simulate and test the above examples, and then predict unknown gene classification data from the model.
Key word artificial neural networks; genetic engineering; exorns; introns