This document summarizes a study that evaluated 8 methods for detecting selection along environmental gradients in populations. Through simulations, the study tested the effectiveness of methods based on population differentiation (FST) and environmental correlation in outbreeding and selfing populations under varying levels of selection strength and sampling. The results showed that methods using environmental data were more powerful at detecting selection, even when selection was weak. Sampling more populations was also found to be most efficient. The document concludes that these correlation-based methods are better able to detect selection, but require good environmental data, and discusses applications to ongoing studies in Medicago truncatula and rice.

1. Detecting Selection Along Environmental
Gradients:
Analysis of Eight Methods and Their Effectiveness for
Outbreeding and Selfing Populations
Yves Vigouroux
Institut de Recherche pour le Développement
Montpellier, France
International Workshop on
“Applied Mathematics and Omics Technologies for Discovering Biodiversity and Genetic Resources for
Climate Change Mitigation and Adaptation to Sustain Agriculture in Drylands”
Rabat - Morocco, 24-27 June 2014

2. Climate change
Rainfall evolution:
1990 - 2090
IPCC 2007 august
Challinor et al.
Rainy season
evolution by 2090

3. Control
crosses or
genealogy
Experimental
population
Natural
population
Analysis of
given traits Association mapping
Association (genealogy,
QTL, NAM, MAGIC)
Selection:
traits?
Genome selection
scan
Segregation
biais
Methodological approaches

4. Principle of differentiation selection
scan
Environment – spatial variation
Neutral allele: variation due
to demographic/gene
flow/history effects
Selected gene: variation due
to demographic/gene
flow/history effects and
selectionDifferentiation
FST

5. • Methods: use or not environmental data?
• Sampling design?
• Impact of the reproduction system (selfing)?
Exemple of
environmental
gradient in
Niger, Africa
Environmental Gradients

6. Software QuantiNEMO [Neuenschwander et al. 2008 Bioinformatics]
Simulations =
- time-forward
- individual
- modèle flexible
 100 populations - 2N = 200
 100 unlinked neutral locus
 1 linked selected locus
 Selfing: {0.0, 0.95}
 Different sampling strategy
Model
si

7. Migration
model

8. Modélisation
sélection
sélection

9. Simulation of selection
• Two allele A and a
• AA fitness 1, Aa fitness 1-s, aa fitness 1-
2s

10. Simulation of selection: example

11. Modélisation: exemple

12. Sampling :
from 1 individual per population to 48.

13. Methods studied
Name Reference Data
FDIST Beaumont et al. 2006 Population
DETSEL Vitalis et al. 2001 Pairs of population
FLK Bonhomme et al. 2010 Population
BAYSCAN Foll et al. 2008 Population
BAYENV Coop et al. 2010 Population
SAM Joost et al. 2006 Individual
GEE Poncet et al. 2010 Individual
Differentiation-
based methods
Correlation
based
methods
Use of
environ.
data

14. Evaluation of the methods
• Simulation of neutral and selected locus
• Use of the method to calculate the
proportion of loci detected
Simulated % of loci detected
selected
Neutral Percentage of false
positive
Expected 5%
Selected Percentage of true
positive
Expected close to 100%

15. ResultsFalsepositiveTruepositive
Correlation to environmental
data based methods
Differentiation (FST) based
method

16. Falsepositive
Correlation to
environmental
data based
methods
Differentiation
(FST) based
method
Results

17. Reproductive system
FalseposiitveTruepositive
Outbreeding Selfing

18. Selection strenght
Method using
environmental data are
more powerful evenif
selection is weaker

19. Sampling
A small number of individual is

20. Conclusion
- Methods based on differentiation are conservative
- Methods based on correlation more powerfull / efficient
- Requiere to have good environmental data...
- New development: non parametric correlation
- Sampling more population is the most efficient
 De Mita et al., 2013, Molecular Ecology
 De Mita et Siol, 2012, BMC Genetics

21. Current application
Genomic analysis
A set of 89 000 SNP
Detection of selection approach
Association mapping

22. Work going on using this approach
Medicago truncatulaRice
(Oriza sp)
Natural populationTraditionnal varieties in Guinea
and Madagascar
Selfing
selfing

23. Acknowledgement
IRD, Montpellier On ongoing project:
S De Mita UAM Niamey
AC Thuillet Y Bakasso
JL Pham IS Ousseini
C Berthouly
CIRAD, Montpellier ISRA Sénégal
N Ahmadi N Kané
INRA Montpellier
L Gay
Université de Provence
S Manel
ARCAD Project
Agropolis Researcher Center for Crop Diversity and Adapation