A Modified Apriori Algorithm for Analysing High-Dimensional Gene Data

Modern high-throughput technologies allow the systematic characterisation of an organism but provide excessive amounts of data such as results from microarray gene expression experiments. Combining the information from various experiments will help to expand the knowledge about an organism. However, the analysis of a data set comprising measurements for thousands of genes under many conditions, requires efficient techniques to be feasible at all. Here, we refine a frequent item-set mining approach for scanning a high-throughput data set in order to identify subsets of genes and subsets of conditions with similar data patterns. As a use case, screenings of 4699 mutant clones of Pseudomonas aeruginosa each with a disrupted gene were considered under 109 conditions. We found an unexpected gene group with highly overlapping phenotypes. Therefore our approach is suitable to simultaneously find objects with similar pattern in high-dimensional data sets and their key characteristics within reasonable time.