The Medicago Pan-Genome Reveals Large-Scale Variation

Traditional read-mapping approaches are inherently limited in repeat-rich genomic regions due to the difficulty in confidently mapping reads to the reference. As a result, variant detection has been hindered in such regions as have studies of complex gene families with high sequence similarities among members. Here we propose a “Pan-genome” approach to investigate the true level of genomic variation through deep sequencing, de novo assembly and comparative analysis. In addition to the Medicago reference genome (HM101), we have sequenced two other Medicago accessions (HM056 and HM340) to high depth with libraries of both short and long insert sizes (300bp and 3000bp). Sequence reads were assembled into scaffolds using the de novo assembler ALLPATHS. Genome-wide pairwise alignments (HM056-HM101, HM340-HM101) were then reconstructed for 332Mb of the HM056 assembly and 296Mb of the HM340 assembly. In aligned genomic regions we found extensive variation that includes single nucleotide polymorphisms, short insertion/deletion plus large structural changes. We identified 2.9M and 7.7M single nucleotide changes in HM056 and HM340 respectively (SNP bp^-1 of 0.87% and 2.62%), levels dramatically higher than previous reports that were based on aligning next generation reads to the reference genome sequence. In an effort to better characterize the genomic content unique to individual Medicago accessions, we identified 7.3 and 7.9 Million bp of novel sequence segments (> 1000bp) in HM056 and HM340 using an ad hoc iterative pipeline.  A “Pan-3” genome was then constructed using the reference genome (HM101) plus the HM056-specific and HM340-specific sequence contents, enabling development of an Agilent CGH microarray for validation. Here, we describe early studies of structural variation detection using the pan-genome approach focusing on the impacts of structural variations on genome architecture, functional relevance to known gene families, and segregating patterns in natural populations.