Order-Preserving SubMatrix (OPSM) has been proved to be important in modelling biologically meaningful subspace cluster, capturing the general tendency of gene expressions across a subset of conditions. Given an OPSM query based on row or column keywords, it is desirable to retrieve OPSMs quickly from a large gene expression dataset or OPSM data via indices. However, the time of OPSM mining from gene expression dataset is long and the volume of OPSM data is huge. In this paper, we investigate the issues of indexing two datasets above and first present a naive solution pfTree by applying prefix-Tree. Due to it is not efficient to search the tree, we give an optimization indexing method pIndex. Different from pfTree, plndex employs row and column header tables to traverse related branches in a bottom-up manner. Further, two pruning rules based on number and order of keywords are introduced. To reduce the number of column keyword candidates on fuzzy queries, we introduce a First Item of keywords roTation method FIT, which reduces it from n! to n. We conduct extensive experiments with real datasets on a single machine, Hadoop and Hama, and the experimental results show the efficiency and scalability of the proposed techniques.
展开▼
