Authors : Keon Myung Lee, Chan Hee Lee and Hyung Woo Youn
Abstract: The advances in molecular technology enable to classify organisms based on their genetic sequence information. In classification, it is sometimes desirable to have a pattern for each class which characterizes the class and discriminates it from others. The objective of this research is to develop a method to extract signature sequences which is a sequential pattern with such characteristics from two contrasting sequence groups. It is assumed that there are two sequence groups, the self group and the other group and all the sequences from both groups are multiply aligned together, so that, they have the same length. To begin with for each group the relative base frequencies at each base location are computed and its consensus sequence is identified. From each base location of a group, the most frequent base is selected as a constituent of signature sequence only when its relative frequency is higher than that of the same base in the corresponding location of the contrasting group by at least the specified threshold called base frequency difference threshold, BDT. A candidate signature sequence is constructed by placing those selected bases in their location of a sequence. A desirable signature sequence for a sequence group is a sequential pattern which facilitates to discriminate its group from the other group and retains its unique group characteristics. In an experiment of virus sequences, the cross-validation study showed that the method produces consistent results and the generated signature sequences give the high sensitivity and high specificity for the sequence data set. The results indicate that the proposed signature sequence extraction method is useful in the characterization and classification of a group of sequences.
Keon Myung Lee, Chan Hee Lee and Hyung Woo Youn, 2018. A Relative Frequency-Based Signature Sequence Extraction Method for Two Contrasting Sequence Groups. Journal of Engineering and Applied Sciences, 13: 3198-3203.