Cancer Outcome Prediction by Cluster-based Artificial Immune Networks

D. Tsankova and V. Rangelova (Bulgaria)


Artificial immune network, hierarchical clustering, gene expression profiling, and diffuse large B-cell lymphoma


Molecular analyses of clinical heterogeneity in diffuse large B-cell lymphoma (DLBCL) focus on individual genes, some of which are correlated with cancer treatment outcome. The treatment depends on the distinction of the significant subtypes of DLBCL and influences on the cured/fatal outcome of the disease. A molecular cured/fatal outcome predictor would be able to determine the proper treatment (chemotherapy or stem cell support) of the cancer and would increase the survival probability. The paper proposes a methodology of designing artificial immune networks for a classification problem oriented to cancer outcome prediction. The immune network structure largely is automatically determined using hierarchical clustering of gene expression data. The proposed cancer outcome predictor is tested in MATLAB environment on the 58 data samples (32 cured and 26 fatal) available in the literature. A high prognostic accuracy (93.1 %) is achieved by a 13-gene outcome prediction model.

Important Links:

Go Back