Application of Artificial Neural Networks in Cancer Classification and Diagnosis Prediction of a Subtype of Lymphoma Based on Gene Expression Profile

Diffuse Large B-cell Lymphoma (DLBCL) is the most common subtype of non-Hodgkin s Lymphoma. DLBCL patients have different survivals after diagnosis. 40% of patients respond well to current therapy and have prolonged survival, whereas the remainders survive less than 5 years. In this study, we have applied artificial neural network to classify patients with DLBCL on the basis of their gene expression profiles. Finally, we have attempted to extract a number of genes that their differential expression were significant in DLBCL subtypes. Methods: We studied 40 patients and 4026 genes. In this study, genes were ranked based on their signal to noise (S/N) ratios. After selecting a suitable threshold, some of them whose ratios were less than the threshold were removed. Then we used PC A for more reducing and Perceptron neural network for classification of these patients. We extracted some appropriate genes based on their prediction ability. Results: We considered various targets for patients classifying. Thus patients were classified based on their 5 years survival with accuracy of 93%, in regard to Alizadeh et al study results with accuracy of 100%, and regarding with their International Prognosis Index (IPI) with accuracy of 89%. Conclusion: Combination of PCA and S/N ratio is an effective method for the reduction of the dimension and neural network is a robust tool for classification of patients according to their gene expression profile.