Bar-coded multiplexed sequencing approaches based on newgeneration sequencing technologies provide capacity to sequence a mapping population in a single sequencing run. However, such approaches usually generate low-coverage and error-prone sequences for each line in a population. Thus, it is a significant challenge to genotype individual lines in a population for linkage map construction based on low-coverage sequences without the availability of high-quality genotype data of the parental lines. In this paper, we report a method for constructing ultrahigh-density linkage maps composed of high-quality single-nucleotide polymorphisms (SNPs) based on low-coverage sequences of recombinant inbred lines. First, all potential SNPs were identified to obtain drafts of parental genotypes using a maximum parsimonious inference of recombination, making maximum use of SNP information found in the entire population. Second, high-quality SNPs were identified by filtering out low-quality ones by permutations involving resampling of windows of SNPs followed by Bayesian inference. Third, lines in the mapping population were genotyped using the high-quality SNPs assisted by a hidden Markov model. With 0.05× genome sequence per line, an ultrahigh-density linkage map composed of bins of high-quality SNPs using 238 recombinant inbred lines derived from a cross between two rice varieties was constructed. Using this map, a quantitative trait locus for grain width (GW5) was localized to its presumed genomic region in a bin of 200 kb, confirming the accuracy and quality of the map. This method is generally applicable in genetic map construction with low-coverage sequence data.genomics | maximum parsimony of recombination | Bayesian inference | hidden Markov model | rice G enetic maps provide the bases for a wide range of genetic and genomic studies and are pivotal for mapping and identifying genes associated with phenotypic performance, referred to as traits. The resolution of a genetic linkage map depends on the number of recombination events in the mapping population and the density of molecular markers. The number of recombination events depends on how the population is created, whereas the density of the markers can be improved continually with advances in molecular techniques. Traditional molecular markers that have been widely used in genotyping assays of populations, although laborious and time-consuming, have limitations in throughput and can provide information only for low-density maps, and thus are of low efficiency.Oligonucleotide microarrays, composed of millions of probes based on genome sequences, can capture large numbers of sequence variations between different samples by comparative genomic hybridization, which can be used for high-throughput marker discovery and genotyping (1-3). However, restrictions in microarray design and the number of probes on the microarrays limit the applications of this technology. In addition, it is cost-prohibitive for genotyping especially if the mapping population is large.New sequencing t...