Bayesian parentage analysis with systematic accountability of genotyping error, missing data and false matching

Abstract

Motivation: The goal of any parentage analysis is to identify as many parent-offspring relationships as possible, while minimizing incorrect assignments. Existing methods can achieve these ends, but they require additional information in the form of demographic data, thousands of markers and/or estimates of genotyping error rates. For many non-model systems, it is simply not practical, cost-effective or logistically feasible to obtain this information. Here, we develop a Bayesian parentage method that only requires the sampled genotypes to account for genotyping error, missing data and false matches. Results: Extensive testing with microsatellite and SNP datasets reveals that our Bayesian parentage method reliably controls for the number of false assignments, irrespective of the genotyping error rate. When the number of loci is limiting, our approach maximizes the number of correct assignments by accounting for the frequencies of shared alleles. Comparisons with exclusion and likelihood-based methods on an empirical salmon dataset revealed that our Bayesian method had the highest ratio of correct to incorrect assignments. Availability: Our program SOLOMON is available as an R package from the CRAN website. SOLOMON comes with a fully functional graphical user interface, requiring no user knowledge about the R programming environment. In addition to performing Bayesian parentage analysis, SOLOMON includes Mendelian exclusion and a priori power analysis modules. Further information and user support can be found at https://sites.google.com/site/parentagemethods/.

Low rates of pollen contamination in a Scots pine seed orchard in Sweden: the exception or the norm?

“…To overcome this problem, the authors proposed relaxing the exclusion criteria, so that a candidate father was excluded only if an offspring mismatched on more than just a single locus. Such a treatment did not cause significant biases in paternity assignments under low mistyping rates; however, Christie et al (2013) showed that relaxing exclusion criteria may be associated with high risk of false assignments, as allowing for only one of eight loci to mismatch led to a false assignment rate of 39%.…”

The genetic approach for assessing sea trout stock enhancement efficiency – An example from the Vistula River

“…In 2015, approximately 2.5 million salmon smolts and 800,000 sea trout smolts were marked with this method in the Baltic basin. In Poland, this marking method has been used in Pomeranian rivers since 2006; however, in response to warnings from veterinarians, this type of marking was halted in 2013and 2014(ICES 2016. Currently, the question has been resolved at least in the Vistula catchment, where this method is not used.…”

“…Methods based on likelihood use quantitative Mendelian inheritance to calculate probability in various candidate relationships within a set, and then the relationships with the greatest inferred probability are selected. The most popular programs with similar algorithms are Cervus (Marshall et al 1998), Colony (Wang 2004), PAPA (Duchesne et al 2002), and SOLOMON (Christie et al 2013). When choosing an appropriate model for parentage analysis, several questions must be answered, most importantly what is the aim of the analysis and what kind of data sets are available.…”

“…1). Assessing genetic polymorphism and kinship in the parent/offspring lines among the sea trout groups analyzed was performed using GenAIEx 6.5 (Peakall and Smouse 2012), FAP (Family Assignment Program) (Taggart 2007) and SOLOMON (Christie et al 2013). FAP is an exclusion-based method program that generates three types of information: (1) the prediction power of a given panel of loci for assigning offspring to parental pairs; (2) the actual assignment of offspring to parental pairs; and (3) the identification of potentially problematic loci.…”

“…To identify true parent-offspring pairs, it determines the posterior probability that the assumed parental pair is false based on the frequency of common alleles. Calculations were performed according to recommendations for microsatellite loci: 1,000 simulated data sets for 50,000,000 simulated genotypes (Christie et al 2013). …”