Sugarcane Genome De Novo Assembly Challenges

Sugarcane is among the most important crops in the world, as both one of the most productive biofuels as well as one of the most important foods. Sugarcane ethanol is a clean, renewable fuel that produces on average 90 percent fewer carbon dioxide emissions than oil and can be an important tool in the fight against climate change. A sugarcane reference genome would facilitate the development of improved strains with higher growth rate, sugar production, disease resistance, and drought resistance. However, the sugarcane genome is very difficult to assemble due to its genomic heritage. One hundred years ago, S. spontaneum, which contributed to robustness, and S. officinarum, which contributed to sweetness were crossed, then the F1 was crossed once again with S. officinarum to further fortify sweetness. Current sugarcane cultivars are highly and variably polyploid, with 10 chromosomes of 8-12 copies each, totaling around 10 Gbp and 100-120 chromosomes. Consequently, they have complicated repeat structure, high aneuploidy, high heterozygosity, along with complex recombination and restructuring of the genome.

Current assembly programs generally target relatively homogenous, diploid genomes such as human, or mouse, and are not well suited for this challenge. Resolving the sugarcane genome requires the analysis and integration of multiple sequencing technologies, and especially long reads. We have generated more than 20x coverage of the genome using Moleculo long read sequencing with read lengths averaging around 5kbp. Adding these data to the assembly using the Celera Assembler improved the contig N50 size by more than 50 fold compared to standard short read Illumina sequencing with a variety of mate pair libraries. We find that most core eukaryotic genes (CEGs) are now well assembled, along with most orthologs from the closely related sorghum genome. We are currently analyzing the assembly graph structure and alignment of multi-platform sequencing data towards a more detailed analysis of the haplotype structure and paralogous gene families. The initial results are promising, particularly due to the low 0.1% error rate of the Moleculo reads, where haplotype information is preserved intact. We validated our results against BAC sequences and against the closely related Sorghum bicolor reference genome.