Rule-based Assembly for Short Read Data Set obtained with Multiple Assemblers and k-mer Sizes

摘要

Various de novo assembly methods based on the idea of k-mer have been proposed. In despite of the success in these methods, another approach, called as Hybrid approach, that combine different traditional methods to take advantages of them have been proposed. However, the results obtained from traditional methods that used in hybrid approach depend on not only the algorithm or heuristics, but also the selection of user-specific k-mer size. Consequently, the results by hybrid approaches also depend on them. In this paper, we designed new assembly approach, called as Rule-based assembly. It follows the strategy similar to the hybrid approach. But it uses certain rules learned from some characteristics of draft contigs to remove erroneous ones and merges them. To construct effective rules, a learning method based on decision tree, called Complex decision tree was proposed. Comparative experiments were also conducted. The results showed that proposed method outperformed traditional one in certain case.