Computational protocol: Human-facilitated metapopulation dynamics in an emerging pest species, Cimex lectularius

Similar protocols

Protocol publication

[…] For each data set, descriptive summary statistics including number of alleles and expected (HE) and observed (HO) heterozygosities were obtained using Microsatellite Analyser version 4.05 (Dieringer & Schlötterer ). Allelic richness was calculated using FSTAT version (Goudet ). We tested for deviations from Hardy–Weinberg equilibrium (HWE) and estimated the frequency of null alleles with cervus version 3.0.3 (Kalinowski et al. ). Evidence of linkage disequilibrium was assessed using genepop version 4.1.0 (Raymond & Rousset ; Rousset ). For analyses of deviation from HWE and evidence of linkage disequilibrium, a Bonferroni correction was applied to account for multiple testing (Rice ). [...] We used FSTAT to calculate global FIS and FST values (Weir & Cockerham ), both within and among infestations. Values were jackknifed over loci to give means and standard errors and bootstrapped over loci to give 95% confidence intervals. Ten thousand permutations were used to generate significance values.The within-city data set was tested for isolation by distance amongst individuals in spagedi version 1.3 (Hardy & Vekemans ) using the kinship coefficient (Loiselle et al. ). Distance was partitioned into 10 intervals, with a uniform number of pairwise comparisons per interval. The mean distance value of each interval was log-transformed (Rousset ). We used 10 000 permutations to test whether the slope of the relationship between geographical and genetic distance was significantly negative.For both data sets, we performed a discriminant analysis of principal components (DAPC; Jombart et al. ) using the adegenet 1.3–4 (Jombart ) package in R (R Core Team ) to examine evidence for genetic clusters, using infestation as a grouping prior. DAPC is an ideal clustering method for this data set as it does not make some commonly required assumptions (e.g. HWE; Jombart et al. ), which are unlikely to hold for bed bug infestations. The first step in DAPC is to transform the raw data into principal components (PCs). There is a trade-off in the number of retained PCs, with a higher number of PCs increasing the ability to discriminate between groups at the cost of the reduced stability of membership probabilities (Jombart et al. ). We used a-score as a measure for judging the optimal number of retained principal components. The a-score is the difference between the proportions of successful observed discriminations and values obtained from random discrimination. This was calculated with 100 permutations for each increasing number of retained principal components using the optim.a.score function in adegenet. Due to the low number of sampled individuals in each group, we were conservative with the number of retained principal components (PCs), but for both data sets, the number of retained PCs still incorporated ≥75% of the variance in the data. The dapc function was then used to perform the clustering analysis, and results are presented as ordination plots. [...] Our aim was to construct simple models that nevertheless allow us to capture what is known of bed bug biology and to retain the key differences between competing hypotheses about colonization dynamics. By keeping the models as simple as possible, they can be assessed with the data available and relevant parameters can be estimated. We note that an important aspect of the bed bug system is the severe bottlenecks experienced by local populations during establishment of new infestations (Doggett et al. ; Saenz et al. ). These successive bottlenecks lead to the loss of most information concerning ancient demographic events, and inferences are thus restricted to the few most recent colonizations. We therefore choose to contrast simplified versions of the ‘propagule pool’ and ‘migrant pool’ models in a coalescent framework rather than trying to simulate colonization, migration and extinction more fully in a classical metapopulation framework. Our models still capture the main difference between these two metapopulation models, and their simplicity allows easier parameter estimation (see below). We used ABC to test the posterior probabilities of the two hypotheses and to estimate parameters for the preferred scenario. To demonstrate the wide applicability of this approach, we conducted all ABC analyses with the readily available and user-friendly software package diyabc (version (Cornuet et al. ). […]

Pipeline specifications

Software tools Cervus, Genepop, SPAGeDi, adegenet, DIYABC
Applications Phylogenetics, Population genetic analysis
Organisms Homo sapiens, Cimex lectularius
Diseases Tick Infestations