Hybrid Method of Using RNA-Seq for Genome Annotation

Usually, when RNA-Seq data is used for a genome annotation the reads are aligned to the reference genome, and these alignments are used as building blocks for the reconstructed gene models. The reads also could be assembled into a set of full-length transcripts using a de novo transcriptome assembler. In our previous work we demonstrated that a typical draft eukaryotic genome may have anywhere from hundreds to thousands genes for which some exons fall into genomic gaps. The reconstructed transcripts of these genes could be dramatically improved if we fill in the missing exons using appropriate parts of the de novo assembled transcripts.

Assembling the full set of RNA-Seq data for a deep transcriptome sequencing project is a computationally expensive process which can yield millions of transcripts. Only a small fraction of these transcripts would be used for the gap-filling of partial genes.  For this purpose we propose a hybrid annotation method in which missing parts of the transcripts are assembled on demand only when gene fragments separated by a genomic gap are found. We show that partial transcripts derived from the genome could be used to locate a subgraph of the general de Bruijn graph. This subgraph is usually small enough to be analyzed fully on the fly and produce one or more transcripts connecting the fragments. This method is evaluated on the baboon and rabbit genomes. The baboon gap-filled transcripts are compared with the human data. The rabbit ones are compared with the transcripts produced by Trinity assembler.