4. Landscape Genomics


8.30am–5.30pm 28 March


ANU Commons Function Centre (Cnr Barry Drive and Marcus Clarke Street, ANU)



Background information and downloads needed for this session can be found here.

The session's notes and slides can be found here.


Session outline:

1. Introduction

  • a. Brief history of Landscape genetics (LG) and its major publications in the last 10 years. 
  • b. Major differences between adaptive LG and population genetics approaches in the detection of neutral and adaptive variation
  • c. Consequences of next generation sequencing datasets on the methods used to detect natural selection

2. Methods

Our course will be focused on correlative approaches. This part will thus show what kind of genetic and environmental data we need and which are the relevant statistic approaches.

a. Sources of environmental data

  • i. Existing sources
    • 1. Climatic variables (CRU, WorldClim, interpolated regional datasets)
    • 2. Satellite imagery, ground cover and soil maps
    • 3. DEM derived variables (Existing, LIDAR and stereophotogrammetry)
  • ii. Fieldwork sources
    • 1. Loggers (temperature, humidity, soil moisture, solar radiation)
    • 2. Indirect ecological indicators
b. Genetic data
  • i. Recoding/Filtering alleles and genotypes
  • ii. Data transformation
  • iii. Note on linkage disequilibrium
c. Statistical approaches 
  • i. Computing associations between genetic data and environmental variables
    • 1. Correlative approaches (GLM) using SAMβada
    • 2. A note on statistical power (false positives)
  • ii. Spatial statistics
    • 1. Concerns regarding spatial autocorrelation. Comparison of results between methods and analysis of spatial dependency between results.
    • 2. LISA 
    • 3. GWR
    • 4. LocalDiff
  • iii. Inclusion of population structure
    • 1. Admixture
    • 2. Correlative approaches including population structure (SGLMM, LFMM, Bayenv)
    • 3. Bivariate models in SAMβada using population structure and environmental variables.

3. Example of studies

  • a. Goats in Morocco (NextGen project)
  • b. Arabis Alpina in the Alps (Intrabiodiv project)
  • c. Biscutella Laevigata at Les Rochers-de-Naye (CH) (local scale)
  • d. Loblolly pine in the US (Eckert et al. 2010)

4. Conclusion and perspectives

  • a. Concluding remark on common findings and differences between datasets
  • b. Challenges in landscape genomics in the coming years

Practical work

1. Acquiring environmental information at sample’s locations
  • a. Choice of a coordinate system
  • b. Retrieving environmental information from major internet sources
  • c. Computing Digital Elevation Model environmental variables
  • d. Extracting environmental variables values at sampling locations
  • e. Exporting data
  • f. PCA to eliminate multi collinearity
2. Identifying loci under selection with SAMβada
  • a. Transforming data from PLINK or others to SAMβada and LFMM
  • b. Parameters of SAMβada
  • c. Console commands
3. Computing membership coefficients and evaluate structure influence using multivariate models.
  • a. Admixture
  • b. LocalDiff
  • c. Multivariate models in SAMβada
4. Identifying loci under selection with approaches including directly population structure
  • a. LFMM
5. Comparative analysis of detected loci
6. Spatial structure analysis of detected loci
  • a. Indicator of spatial autocorrelation using Univariate LISA (Local Indicator of Spatial Association)
  • b. Bivariate LISA of the most relevant associations
7. Genomic position of detected loci and potential gene function
  • a. Comparison with results from (Eckert et al. 2010)
8. Producing results maps in a GIS
  • a. Opening data in Quantum GIS (shape, raster, text delimited)
  • b. Creation of a map (layers, legend, …, output format)


Kevin Leempoel

École Polytechnique Fédérale de Lausanne

Updated:  25 September 2020/Responsible Officer:  Director/Page Contact:  Coordinator