Combining genomic data, new inference methods and long-term population data to uncover population processes

A critical issue in conservation management is to identify the population trajectories of rare and threatened species. Understanding recent population history is central to identifying agents of population decline, and the role processes such as fragmentation and dispersal play in reducing or maintaining population viability. Without understanding these processes, it is difficult to determine how to manage populations of rare species to mitigate adverse impacts and facilitate population recovery. The challenge is that many species of conservation concern are both rare and data deficient, meaning that for most species we have limited understanding of the past population processes needed to inform future management. Traditional ways of inferring population trajectories, such as monitoring change in population size over time, are both costly, time-consuming and often cannot deliver results in the timeframe that conservation managers have to act to avoid further loss.

However, a signature of the past demographic history of a population is contained within the genomes of the individuals currently in that population. This has long been recognised and genetic approaches have been widely used in conservation management to quantify genetic diversity and population structure and to understand processes such as inbreeding and gene flow between populations. The recent and widespread availability of high-quality genomic data, ranging from thousands of loci to populations of whole genomes, coupled with newly developed statistical methods now allow more precise estimates of these demographic parameters. In particular, high-quality genomic data from individuals in a population can be used to reconstruct changes in population size and examine processes such as inbreeding, admixture and selection. This has the potential to fill critical information gaps for species of conservation concern because genomic data can potentially provide high-resolution demographic data from a one-off sampling of individuals in a population. This includes identifying past population trajectories to understand causes of decline, estimating pre-decline levels of genetic diversity thus providing targets for restoring diversity through genetic rescue, captive breeding and translocations, and identifying genes under selection and inferring adaptive potential rather than focusing on neutral genetic diversity1. Genomic data could thus provide managers with critical information to guide conservation actions in a timely and cost-effective manner, leading to a step-change in our capacity to make informed conservation decisions.

Despite this potential, there are substantial challenges to the real-world application of these approaches2, specifically:
1) Methods to reconstruct population size and infer population processes have mostly been based on coalescent approaches, including methods such as site-frequency spectra and sequentially Markovian coalescent methods3. These methods typically ignore factors that are relevant to species of conservation concern, including population structure and processes such as inbreeding and selection. This is problematic because it has been shown that changes in population structure alone can generate signals in genomic data that are characteristic of changes in population size4. Consequently, concomitant changes in population processes have the potential to confound reconstructions of population size from genomic data, potentially limiting the application of this approach to species of conservation concern.

2) There have been few attempts to validate genomic reconstructions using data on recent population changes. Hence, it is unclear how reliably genomic approaches can be used to reconstruct demographic processes on time-scales relevant to conservation management.

We propose to run a series of workshops aimed at overcoming these challenges through combining genomic data (SNPs, whole genome sequences, population haplotyping), new inference methods (see below) and long-term ecological data to critically test assumptions and improve our ability to use genomics evidence to uncover recent population processes.

Simulation is a powerful tool for understanding how concomitant changes in population structure, such as past changes in population size associated with fragmentation, affect contemporary genomic patterns. Researchers at the University of Canberra have developed a comprehensive R package (slimr) that interfaces with the software package SLiM to simulate genomic evolution, facilitating the running and processing of simulations and integration with ecological data. The SLiM software simulates genomic evolution including linkage and selection on a chromosome-wide scale while incorporating potentially complex scenarios of demography and population substructure, different models for selection and dominance, realistic gene structure, and user-defined recombination maps. Such simulation tools provide a way to infer past demographic processes using Approximate Bayesian Computing (ABC) to align simulation results with genomic data to identify the most likely scenarios leading to observed outcomes5. Nevertheless, the potential for these inference methods to uncover population processes on time-scales relevant to conservation management is unclear. The first aim of the workshops will be to evaluate the potential of existing SliM-ABC type approaches to infer recent demographic processes given the availability of different types of data (SNPs, whole genomes, haplotypes). We will use simulation to evaluate how well genomic data can be used to reconstruct population processes of varying complexity, and to identify the data requirements of the inference methods by varying the type of genomic data and sample sizes available.

The second aim of the workshops is to field-validate these approaches for case-study populations where we have existing genomic and long-term population monitoring data (e.g.,6). We will use the genomic data and the inference methods above to reconstruct the most likely trajectory of population change that could have generated the observed genomic patterns and compare the inferred trajectories with the known population history for the chosen case-studies. Reconstructions will use existing knowledge of the ecological processes likely to have influenced the populations, which could be framed as a set of hypotheses, allowing us to evaluate how reliably genomic data can be used to reconstruct recent population history, and the hypothesised processes that need to be incorporated into simulations to achieve reliable reconstructions.

The workshops will bring together researchers working at the interface of genomics and conservation ecology. We propose a series of staged workshops, initially online to prepare the analytical framework, refine the methods and identify and gather relevant population and genomic datasets. We then propose to run two face-to-face workshops to carry out the evaluations and summarise the findings.

The outcome of the workshops will be a data-driven publication that critically evaluates the potential for genomic data coupled with new inference methods to reliably uncover population processes of direct relevance to conservation management. This will substantially improve our ability to determine when and where genomic data have the potential to improve conservation outcomes.