Using Bayesian Statistics to Predict Toxic Response Based on Gene Expression Values

In the emerging field of toxicogenomics, microarray technology is used to read the level of gene expression in cells subjected to various toxic and non toxic agents. The amount of levels that are read from one single array is in the thousands, and when the number of arrays increases, statistical methods are needed to process the information. This master’s thesis documents the use of Bayesian statistics to create two versions of the Bayesian Classifier, both trained to make the distinction between treatment by toxic and non toxic agents based on Affymetrix array data. The classifiers are optimised with regard to the number of expression levels used, and performance is measured by performing various types of cross validation. The first classifier implemented is the Naive Bayesian Classifier, sometimes unjustly called the "idiot’s Bayes". It is proven that the naive classifier works surprisingly well on the noisy and many dimensional array data, in spite of its somewhat naive assumption that all dimensions of the data space are governed by independent probability distributions. The second version of the classifier, which I have chosen to call the "Non-Naive Bayesian Classifier", is designed to take into consideration the dependencies that do exist between the probability distributions of different dimensions of the data space. It does this by creating a graph over the dependencies in the data, based on Mutual Information. However, the Non-Naive Classifier proves to be a disappointment when trying to classify the array data, only increasing the success rates by a few percent. Apart from being used for mere classification, the classifiers are also used to mine the toxicological data, finding relationships between agents and between gene expressions. Clustering finds that some of the agents used belong to a special subgroup, being hard to classify with either version of the classifier. Användning av Bayesisk statistik för att förutsäga toxiskt svar baserat på genuttryck