De novo construction of a "Gene-space" for diploid plant genome rich in repetitive sequences by an iterative Process of Extraction and Assembly of NGS reads (iPEA protocol) with limited computing resources

Abstract

Background: The continuing increase in size and quality of the "short reads" raw data is a significant help for the quality of the assembly obtained through various bioinformatics tools. However, building a reference genome sequence for most plant species remains a significant challenge due to the large number of repeated sequences which are problematic for a whole-genome quality de novo assembly. Furthermore, for most SNP identification approaches in plant genetics and breeding, only the "Gene-space" regions including the promoter, exon and intron sequences are considered.

Results: We developed the iPea protocol to produce a de novo Gene-space assembly by reconstructing, in an iterative way, the non-coding sequence flanking the Unigene cDNA sequence through addition of next-generation DNA-seq data. The approach was elaborated with the large diploid genome of pea (Pisum sativum L.), rich in repetitive sequences. The final Gene-space assembly included 35,400 contigs (97 Mb), covering 88 % of the 40,227 contigs (53.1 Mb) of the PsCam_low-copy Unigen set. Its accuracy was validated by the results of the built GenoPea 13.2 K SNP Array.

Conclusion: The iPEA protocol allows the reconstruction of a Gene-space based from RNA-Seq and DNA-seq data with limited computing resources.

Fig. 1

The evolution of the reconstruction of a genic sequence, represented by a succession of exons (orange) and intron (green). The paired-end reads are colored in orange or in green if they correspond to an intronic sequence (green), an exonic sequence (orange) or a junction intron/exon sequence (green and orange). The figure shows that during the iterations, the introns are added to the exons present in the first reference sequence (Unigene), providing additional information for the next iteration

Fig. 2

The method diagram. For the first iteration, HiSeq 2000 short reads were submitted as “input short read” while longer reads from MiSeq are submitted as “input long reads”. For further iteration Ij, the contigs produced by Ij-1 are used as a new “input long read”, in order to maintain the assembly already produced

Fig. 3

The algorithm of the processing

Fig. 4

During the successive iterations, a the number of de novo genomics contigs decreases, b the mean of the de novo contig length increases, until reaching a plateau at the 6th iteration

Fig. 5

The results of a local BLAST between the 40,227 Unigene contigs allow the estimation of the reconstruction rate of 35,400 de novo contigs at the end of the assembly (iteration I6)