| Gene chips have been used to identify differential gene expression in many areas of biomedical research. cDNA microarray experiments include many distinct stages, such as cDNA purification, printing process, hybridization etc. In each of such stage, various noises and qualities of results often can be introduced. Usually, bad spots with low qualities are deleted in the results of data analysis, and arrays are often excluded from analysis if they include too many bad spots. Such method might not be satisfactory in some situations. In order to extract more valuable data, a more reasonable approach can be developed. The present thesis is mainly concerned with such an approach in the analysis of cDNA microarrays. With bioinformatics and statistical methods, we hope to find some sort of weights to measure the quality of bad microarrays with bad spots. For instance, the more bad spots the microarrays have, the lower weights, and the more good spots the microarrays have, the higher weights. Linear models are used to analyze and assess differential gene expression in this project. The results of the project can show the functions and effects of the weights. Cross-validation is also used to estimate whether using weights is better than not using weights in the linear model. |