In his project I explored various pattern recognition techniques in a biomedical problem. Gene expression profiles are widely studied to classify patients into tumor/non-tumor and/or to the correct tumor category. Since the number of tissue samples examined is usually much smaller than the number of genes examined, efficient data reduction techniques are important. Another interesting problem is to find the most relevant/informative genes in the genome, which can help in successful classification. The aim of this project was to study the effectiveness of linear and non-linear dimensionality reduction methods on performance of classification algorithms by measuring their success rate. In addition, most relevant/discriminative features were found and their success in classification techniques was studied.