In many population genetic problems, parameter estimation is obstructed by an intractable likelihood function. Therefore, approximate estimation methods have been developed, and with growing computational power, sampling-based methods became popular. However, these methods such as Approximate Bayesian Computation (ABC) can be inefficient in high-dimensional problems. This led to the development of more sophisticated iterative estimation methods like particle filters. Here, we propose an alternative approach that is based on stochastic approximation. By moving along a simulated gradient or ascent direction, the algorithm produces a sequence of estimates that eventually converges to the maximum likelihood estimate, given a set of observed summary statistics. This strategy does not sample much from low-likelihood regions of the parameter space, and is fast, even when many summary statistics are involved. We put considerable efforts into providing tuning guidelines that improve the robustness and lead to good performance on problems with high-dimensional summary statistics and a low signal-to-noise ratio. We then investigate the performance of our resulting approach and study its properties in simulations. Finally, we re-estimate parameters describing the demographic history of Bornean and Sumatran orang-utans.

Estimating the age and strength of beneficial alleles is central to understanding how adaptation proceeds in response to changing environmental conditions. Several haplotype‐based estimators exist for inferring the age of segregating beneficial mutations. Here, we develop an approximate Bayesian‐based approach that rather estimates these parameters for fixed beneficial mutations in single populations. We integrate a range of existing diversity, site frequency spectrum, haplotype‐ and linkage disequilibrium‐based summary statistics. We show that for strong selective sweeps on de novo mutations the method can estimate allele age and selection strength even in nonequilibrium demographic scenarios. We extend our approach to models of selection on standing variation, and co‐infer the frequency at which selection began to act upon the mutation. Finally, we apply our method to estimate the age and selection strength of a previously identified mutation underpinning cryptic colour adaptation in a wild deer mouse population, and compare our findings with previously published estimates as well as with geological data pertaining to the presumed shift in selective pressure.

Currently there are a number of coalescent simulation programs that support a wide range of features, such as arbitrary demographic models, migration, and sub structure. Defining the model is done typically with either text files or command line switches. Although this has proven to be a powerful method of defining models of high complexity, it is often error prone and difficult to read without familiarity both with command lines and the program in question. A intuitive GUI based population structure program that can both read and write applicable command lines would dramatically simplify the construction, modification, and error checking of such models by a wider user base.Results: PopPlanner is a tool to both construct and inspect complicated demographic models visually with a GUI where the user’s primary interaction is through mouse gestures. Because of their popularity, we focus on ms and by extension msms, command line coalescent simulation programs. Our program can be used to find errors with existing command lines, or to build original command lines. Furthermore the graphical output supports a number of editing and output features including export of publication quality figures.

Recently, there has been increased awareness of the role of background selection (BGS) in both data analysis and modelling advances. However, BGS is still difficult to take into account because of tractability issues with simulations and difficulty with nonequilibrium demographic models. Often, simple rescaling adjustments of effective population size are used. However, there has been neither a proper characterization of how BGS could bias or shift inference when not properly taken into account, nor a thorough analysis of whether rescaling is a sufficient solution. Here, we carry out extensive simulations with BGS to determine biases and behaviour of demographic inference using an approximate Bayesian approach. We find that results can be positively misleading with significant bias, and describe the parameter space in which BGS models replicate observed neutral nonequilibrium expectations.

In the age of next-generation sequencing, the availability of increasing amounts and quality of data at decreasing cost ought to allow for a better understanding of how natural selection is shaping the genome than ever before. Yet, alternative forces such as demography and background selection obscure the footprints of positive selection that we would like to identify. Here, we illustrate recent developments in this area, and outline a roadmap for improved selection inference. We argue (1) that the development and obligatory use of advanced simulation tools is necessary for improved identification of selected loci, (2) that genomic information from multiple- time points will enhance the power of inference, and (3) that results from experimental evolution should be utilized to better inform population-genomic studies.

The challenge of distinguishing genetic drift from selection remains a central focus of population genetics. Time-sampled data may provide a powerful tool for distinguishing these processes, and we here propose approximate Bayesian, maximum likelihood, and analytical methods for the inference of demography and selection from time course data. Utilizing these novel statistical and computational tools, we evaluate whole-genome datasets of an influenza A H1N1 strain in the presence and absence of oseltamivir (an inhibitor of neuraminidase) collected at thirteen time points. Results reveal a striking consistency amongst the three estimation procedures developed, showing strongly increased selection pressure in the presence of drug treatment. Importantly, these approaches re-identify the known oseltamivir resistance site, successfully validating the approaches used. Enticingly, a number of previously unknown variants have also been identified as being positively selected. Results are interpreted in the light of Fisher's Geometric Model, allowing for a quantification of the increased distance to optimum exerted by the presence of drug, and theoretical predictions regarding the distribution of beneficial fitness effects of contending mutations are empirically tested. Further, given the fit to expectations of the Geometric Model, results suggest the ability to predict certain aspects of viral evolution in response to changing host environments and novel selective pressures.

Genealogies estimated from haplotypic genetic data play a prominent role in various biological disciplines in general and in phylogenetics, population genetics and phylogeography in particular. Several software packages have specifically been developed for the purpose of reconstructing genealogies from closely related, and hence, highly similar haplotype sequence data. Here, we use simulated data sets to test the performance of traditional phylogenetic algorithms, neighbour-joining, maximum parsimony and maximum likelihood in estimating genealogies from nonrecombining haplotypic genetic data. We demonstrate that these methods are suitable for constructing genealogies from sets of closely related DNA sequences with or without migration. As genealogies based on phylogenetic reconstructions are fully resolved, but not necessarily bifurcating, and without reticulations, these approaches outperform widespread 'network' constructing methods. In our simulations of coalescent scenarios involving panmictic, symmetric and asymmetric migration, we found that phylogenetic reconstruction methods performed well, while the statistical parsimony approach as implemented in TCS performed poorly. Overall, parsimony as implemented in the PHYLIP package performed slightly better than other methods. We further point out that we are not making the case that widespread 'network' constructing methods are bad, but that traditional phylogenetic tree finding methods are applicable to haplotypic data and exhibit reasonable performance with respect to accuracy and robustness. We also discuss some of the problems of converting a tree to a haplotype genealogy, in particular that it is nonunique.

BACKGROUND:Anomalous gene trees (AGTs) are gene trees with a topology different from a species tree that are more probable to observe than congruent gene trees. In this paper we propose a rooted triple approach to finding the correct species tree in the presence of AGTs.RESULTS:Based on simulated data we show that our method outperforms the extended majority rule consensus strategy, while still resolving the species tree. Applying both methods to a metazoan data set of 216 genes, we tested whether AGTs substantially interfere with the reconstruction of the metazoan phylogeny.CONCLUSION:Evidence of AGTs was not found in this data set, suggesting that erroneously reconstructed gene trees are the most significant challenge in the reconstruction of phylogenetic relationships among species with current data. The new method does however rule out the erroneous reconstruction of deep or poorly resolved splits in the presence of lineage sorting.

In recent years, methods of approximate parameter estimation have attracted considerable interest in complex problems where exact likelihoods are hard to obtain. In their most basic form, Bayesian methods such as Approximate Bayesian Computation (ABC) involve sampling from the parameter space and keeping those parameters that produce data that fit sufficiently well to the actually observed data. Exploring the whole parameter space, however, makes this approach inefficient in high dimensional problems. This led to the proposal of more sophisticated iterative methods of inference such as particle filters. Here, we propose an alternative approach that is based on stochastic gradient methods and applicable both in a frequentist and a Bayesian setting. By moving along a simulated gradient, the algorithm produces a sequence of estimates that will eventually converge either to the maximum likelihood estimate or to the maximum of the posterior distribution, in each case under a set of observed summary statistics. To avoid reaching only a local maximum, we propose to run the algorithm from a set of random starting values. As good tuning of the algorithm is important, we explored several tuning strategies, and propose a set of guidelines that worked best in our simulations. We investigate the performance of our approach in simulation studies, and also apply the algorithm to two models with intractable likelihood functions. First, we present an application to inference in the context of queuing systems. We also re-analyze population genetic data and estimate parameters describing the demographic history of Sumatran and Bornean orang-utan populations.