A Comparison of Transcript Discovery with Different de novo Assembly Software Tools

Today’s technologies for genome sequencing are faster, cheaper, and creating more sequence data than ever before. In addition, RNA-sequencing (RNA-seq) methods are becoming an increasingly important paradigm in next generation sequencing (NGS) technology to generate transcriptome data from non-model organisms. How to correctly assemble these vast amounts of genomic information for further analyzing, understanding, and make use of them becomes a challenge.

There are many de novo assemblers have been developed to build the sequence of all the transcripts, but there is no significant evidence to show that which assemblers are optimal or perform better than others. We carried out a systematic comparison to show the transcript discovery of 4 different de novo assembler installations. For each assembler, we conducted extensive RNA assemblies with inputs of up to 1.3 billion reads.  We used up to 6 different kmer sizes ranging from one third to the whole read length for each assembly. We also applied the digital normalization, a single-pass computational algorithm that has been claimed to systematize coverage in shotgun sequencing data sets, for input reads pre-processing.

We finally compare the assemblies by looking at the unique and redundant rate across the different combinations of assembliers and kmer sizes.