Genomic research involving gene mapping of economically important traits, population-specific genetic structure and evolutionary history relies heavily on genetic maps built on the extent of linkage disequilibrium (LD) between genomic markers (e.g., Georges et al., 2019; Johnsson and Jungnickel, 2021). For example, to what extent LD persists in a certain genomic region determines the number of markers required to fine-map a quantitative trait loci with succinct power and precision (for review see Qanbari, 2020). Moreover, genetic maps are valuable resources for comparative genomic analyses among breeds or species (e.g., Everts-van der Wind et al., 2005; Womack, 2005). Of utmost topical importance, however, is the contribution of genetic maps (also known as linkage maps) to measuring haplotype similarity in the context of genomic selection (Musa, 2021) and to chromosome-level assemblies of whole-genome sequences (e.g., De los Ríos-Pérez et al., 2020; Rosen et al., 2020).

Given the value of cattle in sustaining the world food security, the bovine genome is subject of vast amount of ongoing research. We recently updated the genetic map of German Holstein breed (Qanbari and Wittenburg, 2020) and compared it with physical coordinates of the most recent bovine reference genome assembly ARS-UCD1.2 (Rosen et al., 2020). As an extension to this resource, here we introduce a Shiny app CLARITY which facilitates interactive visualisation of the bovine genetic and physical map. CLARITY illustrates the details of male recombination across the bovine genome of selected breeds and suggests suitable genetic-map functions. In addition to published findings, results have been updated by taking most recent knowledge about putatively misplaced markers in the bovine genome assembly into account (Qanbari et al., 2022). Moreover, a linkage map for German/Austrian Fleckvieh cattle has been created. The CLARITY app can therefore serve as a toolkit for both educational and research purposes for the genome of bovine and related species.

The app provides access to linkage maps built based on the 50K genotypes of a large pedigree of German Holstein cattle (1,053 half-sib families which comprise 367,056 genotyped animals) as well as German/Austrian Fleckvieh cattle (298,850 genotyped animals pedigreed across 6,866 half-sib families) and based on estimates of recombination rate between intra-chromosomal marker pairs. In what follows, we briefly describe the workflow towards genetic coordinates as depicted in Figure 1 according to Qanbari and Wittenburg (2020). As some of the steps require several days of computing or visual inspection, the workflow was executed once in advance; the app itself dynamically processes physical and genetic coordinates as well as pairwise recombination rates. If not stated otherwise, data processing was executed in R v4.1.3 (R Core Team, 2022).

Step 1:

Genotype data were filtered for minor allele frequency >1% and for Mendelian inconsistencies both on marker and individual level using PLINK v1.9 (Purcell et al., 2007) with recommended settings. Genotypes with a Mendelian inheritance error were set to “NA” and missing values were imputed using Eagle v2.4.1 program (Loh et al., 2016). Putatively misplaced markers, which have recently been reported (Qanbari et al., 2022), were discarded. Data passed on to Step 2 comprised 876 half-sib families with 366,565 progeny genotyped at 44,696 single nucleotide polymorphisms (SNPs) in Holstein and 1,577 half-sib families with 270,636 progeny genotyped at 40,003 SNPs in Fleckvieh.

Step 2:

Paternal recombination rates to build the genetic map were derived from two different methods, see Figure 1. First, the deterministic approach developed by Ferdosi et al. (2014), which is implemented in the R package hsphase v2.0.2, yielded estimates of recombination rate between adjacent markers. These estimates were later used to form the landscape of recombination hotspots. Genetic coordinates were estimated as cumulative sum of recombination rates between neighbouring markers. Second, the likelihood-based approach, as implemented in the R package hsrecombi v0.3.4 (Wittenburg, 2020), was applied to estimate recombination rates between all intra-chromosomal marker pairs. These estimates of recombination rate also enabled the identification of candidate misplaced markers in the current assembly release ARS-UCD1.2 (Qanbari and Wittenburg, 2020). Genetic coordinates were obtained with a smoothing approach, in which all recombination rates < 0.05 between any intra-chromosomal marker pairs were taken into account.

Step 3:

As a novel contribution, the relationship between recombination rate and genetic distance between all intra-chromosomal marker pairs was investigated for a set of commonly used genetic-map functions (Haldane, 1919; Rao et al., 1977; Felsenstein, 1979; Liberman and Karlin, 1984). Given the estimates of recombination rate θ ^ i , j between two markers i and j and its genetic distances d i , j derived in Step 2 (Qanbari and Wittenburg, 2020), a genetic-map function f d i , j | a = θ i , j was fitted to the data by solving the following minimisation problem in terms of the model parameter a (where p is the total number of markers per chromosome): ∑ i , j = 1 i < j p θ ^ i , j − f d i , j | a 2 → min   s . t .   a ∈ R

We solved this optimisation problem using the R function optim with “Brent” option allowing to specify restrictions on a. Rao’s system of mapping function requires a ∈ 0,1 . Furthermore, instead of Haldane’s original map function, we investigated a scaled version thereof, i.e. f d i , j | a = 1 2 1 − e − 2 a d i , j   w i t h   a > 0

For the Binomial map function of Liberman and Karlin (1984), we employed a grid search over a ∈ 2,3,4,5 seeking the minimum squared deviation as described above. The fitted function leading to the least squared deviation constituted the “best” genetic-map function.

The CLARITY app is an R Shiny web GUI for various operating systems. It relies on several R packages to enable the outcome and visualisation functionalities. CLARITY was implemented in R v4.1.3 (R Core Team, 2022) with help of the R packages shiny v1.7.1 (Chang et al., 2021) and shinydashboard v0.7.2 (Chang and Borges Ribeiro, 2021) to create a dashboard. The graphical output was produced using the R packages ggplot2 v3.3.5 (Wickham, 2016), plotly v4.10.0 (Sievert, 2020) and ggVennDiagram v1.2.0 (Gao, 2021). To provide additional helpful features for the app, such as hiding or toogle of elements, the R packages shinyjs v2.1.0 (Attali, 2021) and shinycssloaders v1.0.0 (Sali and Attali, 2020) were incorporated. Tables were generated using the R package DT v0.22 (Xie et al., 2022). Further R packages were employed: cachem v1.0.6 (Chang, 2021), config v0.3.1 (Allaire, 2020), dplyr v1.0.10 (Wickham et al., 2022), gridExtra v2.3 (Auguie, 2017), htmltools v0.5.2 (Cheng et al., 2021), magrittr v2.0.3 (Bache and Wickham, 2022), metathis v1.1.2 (Aden-Buie, 2022), rlang v1.0.6 (Henry and Wickham, 2022a), sf v1.0.8 (Pebesma, 2018), RVenn v1.1.0 (Akyol, 2019) and purrr v0.3.5 (Henry and Wickham, 2022b). Eventually, an R package was built from the CLARITY app with use of the R package golem v0.3.2 (Fay et al., 2022), which offers default R files for creating the package as well as for deploying the app. The R package roxygen2 v7.1.2 (Wickham et al., 2021) was employed for package documentation. The processed data (i.e., recombination rates, genetic coordinates and parameters of genetic-map functions) were included as Rdata files in the folder “extdata”.

We optimized the app following recommendations for best practice with lighthouse (Google LLC, 2022). Figures were compressed with the tool Squoosh (Google Chrome Team, 2022), and caching of those figures requiring longer loading was enabled.

The structure of the app relies on modules, which eases a clear and concise organisation. The use of modules and corresponding interfaces to the main shiny ui and shiny server enables a straightforward maintenance of the software. Furthermore, since each module is an independent app with its own interface and server (Di Filippo et al., 2019), future modification and extension of the app are supported.

The app has three sidebar menus: “Information”, “Breed analysis” and “Breed comparison” which are described in the following.

The CLARITY app provides an environment to interactively explore the physical and genetic map in selected cattle breeds. Importantly, processing of genotype data and presenting results via Shiny app rely on the current bovine genome assembly ARS-UCD1.2. In case of new assembly releases, Steps 1–3 need to be re-run and an app update becomes necessary. (Especially, both approaches in Step 2 depend on the ordering of markers for inferring phases of sire genotypes and recombination events.) Though a pipeline for Steps 1–3 is available at github, and it could theoretically be part of the Shiny app, we do not recommend its inclusion for computational matters as mentioned in Section 2.

Further work on the integration of data from other breeds (beef, dairy, dual-purpose) is underway and will facilitate complex comparative analyses of map features (e.g., hotspot intervals and assembly flaws) in different genomes. Our Shiny app will be extended accordingly, certainly increasing its value for educational and research purposes.