Clustering of Short Read Sequences for de novo Transcriptome Assembly

Abstract

Given the importance of transcriptome analysis in various biological studies and considering the<br />vast amount of whole transcriptome sequencing data, it seems necessary to develop an<br />algorithm to assemble transcriptome data. In this study we propose an algorithm for<br />transcriptome assembly in the absence of a reference genome. First, the contiguous sequences<br />are generated using de Bruijn graph with different k-mer lengths. Then, the eclectic mixtures of<br />sequences are gathered in order to form the final sequences. Lastly, the contiguous sequences<br />are clustered and the isoform groups are provided. This proposed algorithm is capable of<br />generating long contiguous sequences and accurately clustering them into isoform groups.To<br />evaluate our algorithm, we applied it to a simulated RNA-seq dataset of rat transcriptome and a<br />real RNA-seq experiment of the loricaria gr. cataphracta transcriptome. The correctness of the<br />assembled contigs was more than 95%, and our algorithm was able to reconstruct over 70% of<br />the transcripts at more than 80% of the transcripts' lengths. This study demonstrates that<br />applying a sophisticated merging method improves transcriptome assembly. The source code is<br />available upon request by contacting the corresponding author by email.<br />