dos.5.step one PHG imputation reliability for WGS
WGS data for the Chibas founder taxa were downsampled with seqtk (Li, 2013 ) to 1x, 0.1x, and 0.01x coverage. Sequences were produced with three separate seed integers to create three unique sets of reads at each level of coverage. The full WGS data and each set of down-sampled sequencing reads were run through the PHG findPaths pipeline using a PHG database with nodes built from the Chibas founders, minReads = 0, minTaxa = 1, and all other parameters left at default values. Setting the minReads parameter to 0 means that the HMM will attempt to find a path through the entire genome, even when there is no sequence data observed at a particular reference range. Setting the minTaxa parameter to 1 means that all haplotypes are kept, even if taxa are too divergent to group with other individuals in the database. The SNPs were written at all variant sites in the graph, as well as all positions in the sorghum hapmap (Lozano et al., 2019 ). The SNP calling accuracy was assessed by comparing PHG SNP calls to a set of 3,468 GBS SNPs (Muleta et al., unpublished data, 2019). The SNPs with minor allele frequency <.05 or call rate <.8 were removed before comparing PHG and GBS SNP calls. Haplotype calling accuracy was evaluated by running low-coverage sequence through the database and counting the number of times that the selected node in the graph contained the taxon being imputed.
If you find yourself mistake rates for many taxa had been similar to the full error, BF-95-11-195 endured out as which have an excellent four-flex higher mistake than expected in contacting SNPs, whether or not the haplotype getting in touch with mistake wasn’t unusually highest. We suspect that it try is actually mixed up or contaminated that have DNA out-of other shot through the sequencing however, leftover BF-95-11-195 regarding database and you may included it throughout analyses.
2.5.2 Beagle 5.0 imputation precision
Given that PHG is expected to get beneficial whenever merely skim sequence info is designed for one, i opposed PHG imputation reliability so you can Beagle 5.0 (Browning & Browning, 2016 ) imputation accuracy from lower-exposure sequence. The newest WGS research for every taxon is down-sampled as the described over. Each off-sampled dataset and complete-exposure (?8x) WGS research from twenty four creators of your own Chibas sorghum breeding program is actually aimed into sorghum v3.0 resource genome that have BWA MEM (Li & Durbin, 2009 ; McCormick ainsi que al., 2017 ) and you may alternatives were entitled into the Sentieon DNASeq variant contacting pipeline (Sentieon DNAseq, 2018 ). The new VCF data files per inventor was blended playing with bcftools (Li mais aussi al., 2009 ). Whenever version internet failed to fall into line about full coverage WGS (i.elizabeth., a version is actually expected anyone yet not for another such that consolidating variant phone calls round the taxa carry out develop a lost call in specific taxa and you will a different allele get in touch with someone else), the brand new unobserved web site is presumed to-be the new reference call. To help you clear up both Beagle and PHG imputation pipes and since some body included in the databases framework was anticipated to getting inbred contours, the heterozygous phone calls was basically assumed ahead regarding sequencing and you can genotyping problems in the place of recurring heterozygosity and you may was removed. To your down-tested datasets, unobserved web sites were remaining since lost. A reference panel made out of full-coverage WGS was used in order to impute SNPs regarding down-sampled VCF data files. Zero web sites regarding down-sampled data was disguised; alternatively, forgotten information is imputed privately making use of the source panel. On the full-visibility dataset, 1% of all the websites had been masked and you will re also-imputed. Imputation reliability after all amounts of sequence publicity try analyzed from the comparing Beagle phone calls to help you a couple of 3,849 GBS SNPs.