Long-read sequencing of 111 rice genomes reveals significantly larger pan-genomes

Abstract

The concept of pan-genome, which is the collection of all genomes from a population, has shown a great potential in genomics study, especially for crop sciences. The rice pan-genome constructed from the second-generation sequencing (SGS) data is about 270 Mb larger than Nipponbare, the rice reference genome (NipRG), but it is still disadvantaged by incompleteness and loss of genomic contexts. The third-generation sequencing (TGS) with long reads can help to construct better pan-genomes. In this paper, we report a high-quality rice pan-genome construction method by introducing a series of new steps to deal with the long-read data, including unmapped sequence block filtering, redundancy removing, and sequence block elongating. Compared to NipRG, the long-read sequencing-based pan-genome constructed from 105 rice accessions, which contains 604 Mb novel sequences, is much more comprehensive than the one constructed from ∼3000 rice genomes sequenced with short reads. The repetitive sequences are the main components of novel sequences, which partially explain the differences between the pan-genomes based on TGS and SGS. Adding six wild rice accessions, there are about 879 Mb novel sequences and 19,000 novel genes in the rice pan-genome in total. In addition, we have created high-quality reference genomes for all representative rice populations, including five gapless reference genomes. This study has made significant progress in our understanding of the rice pan-genome, and this pan-genome construction method for long-read data can be applied to accelerate a broad range of genomics studies.