Background: The growing wealth of public available gene expression data has made the systemic studies of how genes interact in a cell become more feasible. Liquid association (LA) describes the extent to which coexpression of two genes may vary based on the expression level of a third gene (the controller gene). However, genome-wide application has been difficult and resource-intensive. We propose a new screening algorithm for more efficient processing of LA estimation on a genome-wide scale and apply its use to a data set. Results: On a test subset of the data, the fast screening algorithm achieved >99.8 agreement with the exhaustive search of LA values, while reduced run time by 81-93 . Using a well-known yeast cell-cycle data set with 6,178 genes, we identified triplet combinations with significantly large LA values. In an exploratory gene set enrichment analysis, the top terms for the controller genes in these triplets with large LA values are involved in some of the most fundamental processes in yeast such as energy regulation, transportation, and sporulation. Conclusion: In summary, in this paper we propose a novel, efficient algorithm to explore LA on a genome-wide scale and identified triplets of interest in cell cycle pathways using the proposed method in a yeast data set. A software package named for implementing the algorithm is available through .
Bibliographical noteFunding Information:
The authors are thankful for the resources from the University of Minnesota Supercomputing Institute. The authors are thankful for the helpful discussion with Dr. Jeffrey Leek. Yen-Yi Ho is partially supported by grants 2P30CA077598, P50CA101955, UL1TR000114 and U54-MD008620.
© Gunderson and Ho; licensee BioMed Central Ltd.
- Coexpression pattern
- Genome-wide search
- Liquid association