Abstract
Large repeat rich genomes present challenges for assembly and identification of gene models with short read technologies. Here we present a method we call Virtual Genome Walking which uses an iterative assembly approach to first identify exons from de-novo assembled transcripts and assemble whole genome reads against each exon. This process is iterated allowing the extension of exons. These linked assemblies are refined to generate gene models including upstream and downstream genomic sequence as well as intronic sequence. We test this method using a 20X genomic read set for the axolotl, the genome of which is estimated to be 30 Gb in size. These reads were previously reported to be effectively impossible to assemble. Here we provide almost 1 Gb of assembled sequence describing over 19,000 gene models for the axolotl. Gene models stop assembling either due to localised low coverage in the genomic reads, or the presence of repeats. We validate our observations by comparison with previously published axolotl bacterial artificial chromosome (BAC) sequences. In addition we analysed axolotl intron length, intron-exon structure, repeat content and synteny. These gene-models, sequences and annotations are freely available for download from https://tinyurl.com/y8gydc6n. The software pipeline including a docker image is available from https://github.com/LooseLab/iterassemble. These methods will increase the value of low coverage sequencing of understudied model systems.