A fuzzy rule based approach to identify biomarkers for diagnostic classification of cancers

An important problem for doctors is to identify a small set of useful biomarkers (not all related genes) that can discriminate between different subgroups of cancers which appear similar in routine histology. Here we propose a method for simultaneous feature/gene selection and rule generation for the same problem. Since the feature selection method is integrated into the rule base tuning, it can account for possible subtle nonlinear interaction between features as well as that between features and the tool, and hence can identify a useful set of features for the task at hand. We applied our method to find biomarkers for a group of four childhood cancers that is collectively known as small round blue cell tumors. For this data set first we have used a neural network to reduce the dimension of the data and then applied our method to find biomarkers and rules. Our system could find only eight genes including a novel gene that can do the diagnostic prediction task with a high accuracy. The system can be extended to non-classification applications also.