Associating gene functional groups with multiple clinical conditions using Jaccard similarity

Gene expression arrays provide a rich source of information on the behaviour of thousands of genes for several clinical conditions in a particular tumor/cancer. Such expression sets when integrated with functional classification of genes enrich information provided from both sources. Stemming from the need to score relations between functional groups of genes and multiple clinical types associated with a tumor, this study proposes to use Jaccard similarity. For any set of genes, this measure can be used to measure the association between two sets of gene classes/groups, obtained from two different sources of information. In the proposed study, we particularly consider subsets of overexpressing genes in cancer expression sets. This enables the identification of unique genes and associate their most correlated sample clinical types to their functional groups. Experiments on a breast cancer expression set are done to illustrate the use of the proposed measure.