Efficient amplicon panels for evaluating hybridization and genetic structure in the Schistosoma haematobium species group

Abstract

Genetic markers for detecting hybridization and measuring population genetic parameters must be informative and cost-effective. Most population genetic studies within the Schistosoma haematobium species group rely on either a two-marker system consisting of the mitochondrial, cytochrome oxidase 1 (cox1) and the nuclear internal transcribed spacer (ITS) markers or, at the other extreme, millions of single nucleotide variants (SNVs) from whole genome/exome sequencing. cox1 and ITS studies contain minimal population genetic information but whole genome sequencing is cost-prohibitive. We examined ~38 million previously published, whole genome SNVs genotyped in 162 S. haematobium and S. bovis sampled across Africa. We compared population genetic parameters from 4,000 panels of 10-100,000 randomly sampled SNVs to results from the whole genome dataset to test the resolution of reduced representation sequencing in schistosomes. We found that panels of 500 SNVs captured >99% of the population genetic information contained in the whole genome dataset by using Procrustes transformed principal component analyses and ancestry estimates ( $r^2=0.85$ ). Additionally, the costs of genotyping parasites with an amplicon panel is two to three times less than whole genome sequencing. Our results show that moderately-sized amplicon panels targeting random SNVs provide an efficient approach to large scale, field-based schistosome surveillance.

Keywords: Schistosoma; amplicon panel; hybridization; introgression; single nucleotide variants.

Figure 1.. Similarity between randomly sampled SNV panels and a whole genome dataset.

(A). Procrustes similarity ($t_0$) between panels of random SNVs and the whole genome dataset were calculated across >100 replicate PCAs for panels ranging in size from 10–1,000 SNVs. SNV datasets >1,000 SNVs aren’t shown but show $t_0<0.999$ in all replicates. Red lines indicate mean values, box plots cover the interquartile range from 25%-75% of values. Whiskers indicate the full range of values. All $t_0$ values are shown with grey markers. The source data shown in this figure is available in the supplemental materials.

Figure 2.. Principal component analyses (PCA) converge with increase numbers of SNVs -

PCAs were generated from all filtered variants (n = 381,501 SNVs; red markers) and randomly selected variants (blue markers). Procrustes transformations were used to measure similarity (${\mathit{t}}_0$) between PCAs generated from all and a random subset of variants. We examined SNVs panels containing 10 – 100,00. Results are shown for (A) 10, (B) 20, (C) 50, (D) 100 (E) 200, (F) 300, (G) 400, (H) 500, (I) 1,000, (J) 2,000, (K) 10,000, and (L) 100,000 SNVs at which point the similarity between the panel and full data sets are effectively identical. The source data shown in this figure is available in the supplemental materials.

Figure 3. Comparing ancestry estimates SNV panels and whole genome data.

(A). S. bovis introgressed alleles are present in many S. haematobium individuals. We compared the percent minor parent ancestry ($q$) from SNV panels to estimates made from whole genome data. The two are positively correlated once 50 SNVs are sampled though the strength of the correlation continues to increase. The scatter plots only show S. haematobium individuals. These results show that the panels of SNVs can accurately replicate a set of whole genome SNVs with a few hundred to thousand markers. The source data shown in this figure is available in the supplemental materials.

Figure 4.. Fixation index (FST) versus panel size.

The global, weighted Weir-Cockerham FST was measured between northern S. haematobium, southern S. haematobium, and S. bovis with panels ranging from 10–100,000 SNVs. The red line indicates FST measured with all autosomal, unlinked, common (minor allele frequency >0.05) SNVs in the whole genome dataset (n_SNVs = 381,501). Yellow boxes represent mean values for each panel size. Individual FST values are shown with black dots. Mean Fst is relatively consistent regardless of panel size, but the variation among replicates is inversely correlated the number of makers. The source data shown in this figure is available in the supplemental materials.