Research On Information Processing Methods For Gene-Prediction

The tremendous interest in bioinformatics, a new discipline at the intersection of molecular biology, computer science and mathematic, is fueled by the excitement surrounding the sequcnecing of the human genome and the promise of a new era in which genomic research dramatically improves the human condition. Gene prediction is a vital research field in bioinformatics. And the prediction of the 5'exons is the key and essential point in gene prediction, whose research result not only has great significance in establishing a complete gene prediction mode, but also guides the research of gene expression, gene control, gene functional prediction and biomedical engineering.This dissertation focuses on problems existing in the prediction of 5'exons and explores gene areas especially the signal characteristics in 5'exons area by combining and adopting some biological theories and statistic methods to extract the sequence characteristics modes reflecting the features in this area and make predictions of gene sequence by classifying. At the same time, according to the goals and characterisitics of the issues in predicting 5' exons area, a relatively effective gene predicting algorithm is designed and a gene prediction model is constructed which is based on the statistical combination and the feature classification that improves the predicting precision of of gene characteristics, gene-prediction and integratice prediction. The main research work of this dissertation can be summarized as follows:(1) On the basis of the existing research work of gene prediction, a general frame of 5'exons prediction is designed. This frame aiming at the characteristics of 5'exons, classifies sequences according to gene local features, designs relevant predicting algorithms in terms of features of different types, and then adopts a statistical combination of various information processing methods to predict 5'exons. At last, a predicting model of 5'exons based on the statistical combination and the feature classification is constructed on the basis of this frame.(2) Starting from biological theories, and after the analysis of the three local features of 5'exons-CpG island, TATA-box and DPE by using statistic methods, a predicting method based on gene characteristics by classifying is put forward. The advantage of this method lies in its capability of constructing relevant gene predicting programs according to different geng sequences to make predictions. These predicting programs are more goal-oriented in the constructing process and able to effectively improve the predicting precise of the algorithm.(3) From the angle of informatics, a multi-fractal spectrum is done on gene sequence to gain the characteristics of gene sequence in 5'exons area. And then according to these characteristics an analysis of the difficult points in predicting 5'exons is made as well as the advantages of genetic algorithm in predicting 5'exons. Aiming to the weakpoints of genetic algorithms in maintaining variety mechanism in gene prediction, an immune genetic algorithm is brought forward to predict 5'exons by introducing immune response mechanism and immune network regulation...