A new method of finding groups of coexpressed genes and conditions of coexpression

Abstract

Background: To study a biological phenomenon such as finding mechanism of disease, common methodology is to generate the microarray data in different relevant conditions and find groups of genes co-expressed across conditions from such data. These groups might enable us to find biological processes involved in a disease condition. However, more detailed understanding can be made when information of a biological process associated with a particular condition is obtained from the data. Many algorithms are available which finds groups of co-expressed genes and associated conditions of co-expression that can help finding processes associated with particular condition. However, these algorithms depend on different input parameters for generating groups. For real datasets, it is difficult to use these algorithms due to unknown values of these parameters. Results: We present here an algorithm, clustered groups, which finds groups of co-expressed genes and conditions of co-expression with minimal input from user. We used random datasets to derive a cutoff on the basis of which we filtered the resultant groups and showed that this can improve the relevance of obtained groups. We showed that the proposed algorithm performs better than other known algorithms on both real and synthetic datasets. We have also shown its application on a temporal microarray dataset by extracting biclusters and biological information hidden in those biclusters. Conclusions: Clustered groups is an algorithm which finds groups of co-expressed genes and conditions of co-expression using only a single parameter. We have shown that it works better than other existing algorithms. It can be used to find these groups in different data types such as microarray, proteomics, metabolomics etc.