Haplotype Reconstruction Using Linked-Reads

Haplotype resolution of genomes provides more information for understanding a genome. However, traditional short read methods while scalable, cannot achieve this resolution and provide only a partial view resulting in an underestimation of the biological complexity present. We developed a technology that retains long-range information while maintaining the power and accuracy of short read sequencing. The 10X Genomics Chromium Genome solution starts with only 1ng of DNA and uses haplotype-level dilution of high molecular weight DNA to create a novel data type referred to as “Linked-Reads”.

Coupling Linked-Reads with novel algorithms that take advantage of these linkages allows for improved reference based individual genotype reconstruction without sacrificing variant calling accuracy. Using the NA12878 genome, we construct large phase blocks with an N50 of 4.1 Mb for genomes and 222 Kb for exome. Additionally, Linked-Reads enable the detection of large scale structural variants that are typically challenging for short read methods. We estimate that we can rescue ~50 Mb of previously inaccessible sequence by this method. We identified 2.07X more novel SNPs on average in degenerate regions when compared with standard short read sequencing.

We successfully assembled various non-human genomes including hummingbirds, dogs and chili pepper (respectively 1, 2.5 and 3.5 estimated Gb). Preliminary analysis revealed that the N50 phase blocks were ranging from 0.2 Mb to 11.9 Mb, with N50 contig between 65 and 140 kb. In summary haplotype reconstruction enables genome-wide phasing, rescue of reads previously unable to be aligned, structural variant detection and de novo assembly.