T-IDBA: A de novo Iterative de Bruijn Graph Assembler for Transcriptome

RNA sequencing based on next-generation sequencing technology is useful for analyzing transcriptomes, discovering novel genes and studying exon/intron structures. Similar to genome assembly, de novo transcriptome assembly does not rely on a reference genome and additional annotated information. Most, if not all, existing de novo transcriptome assemblers rely heavily on de novo genome assembly techniques without fully utilizing the properties of transcriptomes and may result in short contigs because of the splicing nature (shared exons) of the genes and the repeats that exist in different genes. In this paper, we analyze the properties of the mammalian transcriptome and propose an algorithm to reconstruct expressed isoforms without a reference genome. We extend the iterative de Bruijn graph approach (IDBA) by using pair-end information to solve the problem of long repeats in different genes and the problem of branching in the same gene due to alternative splicing. The graph will be decomposed into small components, each of which corresponds to a few, if not single, genes. The most possible isoforms with sufficient support from the pair-end reads will be found heuristically by depth-first search. In practice, our de novo transcriptome assembler, T-IDBA, outperforms Abyss (one of the newest de novo transcriptome assembler) substantially in terms of sensitivity and precision for both simulated and real data. The experimental results also match with our theoretical analysis of the performance of T-IDBA, which guarantees most isoforms can be reconstructed as long as their coverage exceeds a certain threshold. Availability: T-IDBA is available at http://www.cs.hku.hk/~alse/idba/