Identification Of Regulatory Patterns Of Amp-Activated Protein Kinase By Association Rules

As an important protein kinase, AMP-activated protein kinase (AMPK) plays a center regulatory role in energy metabolism of cells, and may become an attractive pharmacological target for metabolic diseases, such as obesity and diabetes. Thus, the deep investigation into the structure, expression and activation of AMPK will assist in understanding the regulatory functions and the role to disease states. This could provide an efficient scheme for diagnosis and treatment of human diseases.This paper applies the positive association rules and negative association rule rules that represent different patterns to identify the correlations between the subunit isoforms of AMPK, and between the isoforms and the stimulus factors. The experimental data are collected by searching the relevant bio/medical literatures from NCBI database.Traditional Apriori algorithm usually generates a large number of itemsets and can not efficiently identify interesting ones. Thus, this paper applies several methods to adapt Apriori algorithm. Support constraints are used to extract interesting frequent itemsets. Sorted closure schema is employed to remove the infrequent itemsets from the set of candidate itemsets. Further, ε-cluster is applied to cluster the frequent itemsets and reduce the redundancy of itemsets. Combined with the constraints of itemsets and rules derived from the aim of study, the presented algorithm is used to identify the positive regulation patterns of AMPK. The experimental results demonstrate that the adapted algorithm not only improves the efficiency of Apriori algorithm, but also identifies important activation pattern of AMPK.Like the positive association rules, the negative association rule algorithm also generates vast negative itemsets. To solve this issue, the existing negative association rule algorithm is adapted to effectively extract inhibitory regulation patterns of AMPK. Mutual information is used to capture the items with strong dependence ahead of data mining. Further, another specification of support constraints is applied to control the generation of negative itemsets. The results indicate the proposed algorithm can efficiently identify biologically important inhibitory patterns of AMPK.