Molecular Ecology (2008) 17, 4500–4511 doi: 10.1111/j.1365-294X.2008.03925.x © 2008 The Authors Journal compilation © 2008 Blackwell Publishing Ltd Blackwell Publishing Ltd Pedigree reconstruction in wild cichlid fish populations MARTIN KOCH,* JARROD D. HADFIELD,† KRISTINA M. SEFC* and CHRISTIAN STURMBAUER* *Department of Zoology, University of Graz, Universitätsplatz 2, 8010 Graz, Austria, †Institute of Evolutionary Biology, University of Edinburgh, West Mains Road, Edinburgh EH9 3JT, UK Abstract It is common practice to use microsatellites to detect parents and their offspring in wild and captive populations, in order to reconstruct a pedigree. However, correct inference is often constrained by a number of factors, including the absence of demographic data and ignorance regarding the completeness of parental sampling. Here we present a new Bayesian estimator that simultaneously estimates the pedigree and the size of the unsampled population. The method is robust to genotyping error, and can estimate pedigrees in the absence of demographic data. Using a large-scale microsatellite assay in four wild cichlid fish populations of Lake Tanganyika (1000 individuals in total), we assess the performance of the Bayesian estimator against the most popular assignment program, Cervus. We found small but significant pedigrees in each of the tested populations using the Bayesian procedure, but Cervus had very high type I error rates when the size of the unsampled population was assumed to be lower than what it was. The need of pedigree relationships to infer adaptive processes in natural populations places strong constraints on sampling design and identification of multigenerational pedigrees in natural populations. Keywords: Bayesian parentage assignment, likelihood-based paternity testing, microsatellites Received 19 May 2008; revision received 21 July 2008; accepted 3 August 2008 Introduction Increasingly in evolutionary biology, molecular genetic data are being used to infer the relatedness of a set of sampled individuals. Although methods have been developed for determining relatedness without explicitly considering the underlying pedigree (e.g. Queller & Goodnight 1989), treating the problem of relatedness inference as a problem in pedigree reconstruction is likely to be more efficient (Thomas 2005). However, while pedigrees are often easy to reconstruct in captive or seminatural populations, such relationships are often more difficult to infer in large natural populations (Wilson & Ferguson 2002; Theriault et al. 2007; Pemberton 2008), particularly when behavioural and/or demographic data are unavailable. Generally, pedigree reconstruction methods fall into two groups; those that try to assign offspring to their parents (reviewed in Jones & Arden 2003), and those that try to partition a group of individuals into sibships (reviewed in Blouin 2003). These methods are usually limited to the reconstruction of two generation pedigrees, although parentage assignment methods are often used recursively to reconstruct multigenerational pedigrees when appropriate demographic data have been collected (Pemberton 2008). Although straightforward in principle, such analyses are usually complicated by (i) incomplete sampling of potential parents, (ii) individuals being related to each other in ways not explicitly considered (e.g. individuals belonging to a sibship but only considered as potential parent and offspring), and (iii) non-Mendelian transmission of genotypes (through null alleles, mutations or genotyping error). In addition, when the ages of the sampled individuals are unknown, naïve application of parentage assignment methods can lead to problems where illegal pedigrees are proposed. For example, assigning parentage to a group of individuals from within the same group is likely to result in proposed pedigrees where an individual is its own grandparent. A variety of methodologies currently exist for addressing many of these problems, although there have been few empirical studies comparing and validating the different approaches (e.g. Butler et al. 2004; Csillery et al. 2006; Berger-Wolf et al. 2007). In this study, we sampled almost 1000 individuals from four wild populations of the East African cichlid Tropheus moorii with the aim of assigning individuals to parent–offspring pairs. Using 10 polymorphic Correspondence: Christian Sturmbauer, Fax: 0043 380 9875; E-mail: christian.sturmbauer@uni-graz.at