DNA microarray has gained an increasing popularity for the experimental study of gene-expression for cancer detection. The selection of relevant genes is very important for medical diagnosis as well as to understand how a genome as a whole works for a particular disease. We have proposed a Co-operative Multiswarm Particle Swam Optimization (CMPSO) based technique to find co-regulated gene groups associated with cancer sample categories. The proposed CMPSO is applied to pediatric leukemia dataset and the result is compared with conventional techniques. Proposed multiswarm based technique generates more precise groups of genes which provide important contribution for pediatric leukemia detection. © 2015 Elsevier Ltd. All rights reserved.