D-Mine: Accurate Discovery of Large Pattern Sequences from Biological Datasets

Abstract

Exploring interesting associations on gene variables help to assess the accuracy of the pattern sequence mining. Exploration of genetic structures like DNA, RNA, and protein sequences from biological datasets will boost up new innovations in Pathology diagnosis. For this mission, very large genetic pattern sequences are to be discovered. To do this, doubleton pattern mining (DPM) is considered as very constructive for analyzing these datasets. In this paper, D-Mine, a new approach for discovering very large gene pattern sequences from Biological datasets is discussed. D-Mine effectively discovers doubleton patterns which are further enriched to generate gene pattern sequences with vector intersection operator and Markov probabilistic grammars. D-Mine is described as a solution to diminish the set of discovered patterns. D-Mine makes use of a new integrated data structure called ‘D-struct,’ as combination of a virtual data matrix and 1D array pair set to dynamically discover doubleton patterns from biological datasets. D-struct has a diverse feature to facilitate which is that it has extremely limited and accurately predictable main memory and runs very quickly in memory-based constraints. The algorithm is designed in such a way that it takes only one scan over the database to discover large gene pattern sequences by iteratively enumerating D-struct matrix. The empirical analysis on D-Mine shows that the proposed approach attains a better mining efficiency on various biological datasets and outperforms with CARPENTER in different settings. The performance of D-Mine on biological data set is also assessed with accuracy and F-measure.