Modified LRphase and Simulated Dataset for "LRphase: an efficient algorithm for assigning haplotypic identity to long reads"

A modified version of LRphase was used to simulate reads for a single hypothetical human genome with maternal and paternal phasing information. Briefly, haplotype-specific reference sequences were generated with `bcftools consensus` (Li 2011) based on the rescued GIAB VCF and hg38 human reference sequence. Each haplotype-specific fasta was fed separately into pbsim2 (Ono, Asai, and Hamada 2021) as the reference from which simulated reads were randomly drawn, up to 1X coverage. Parameters controlling the read length distribution, sequencing, and base calling error rates were set to emulate typical performance of the MinIon sequencing platform with flow cell version R10.4.1 (https://nanoporetech.com/products/minion). These are as follows: `--depth 1 –hmm_model R103.model --difference-ratio '23:31:46' --length-mean 25000 --length-min 100 --length-max 1000000 –length-sd 20000 --accuracy-mean 0.98 --accuracy-min 0.01 --accuracy-max 1.00`. Simulated reads were aligned to the hg38 reference genome with minimap2 (Li 2018) and correct phasing and alignment coordinates were encoded in the read names. Finally, samtools (Li et al. 2009) was used to remove duplicated and supplementary reads, and concatenate, sort, and index reads into a single combined bam file. Of 258,539 total reads, 246,210 were mappable, and 178,504 overlapped at least one heterozygous variant in HG001.