Interview with the authors: Background selection and FST: Consequences for detecting local adaptation

Recent work has suggested that background selection (BGS) may lead to incorrect inferences in FST outlier studies, generating substantial concern given the prevalence of these studies in evolutionary biology. In their recent Molecular Ecology publication, Matthey‐Doret and Whitlock investigate the effects of BGS on FST outlier tests using biologically realistic simulations, and find minimal effects. Matthey-Doret and Whitlock suggest that previous studies used unrealistic parameter values in simulations, leading to an overestimate of the effects of BGS in real studies. Read the full article here: https://onlinelibrary.wiley.com/doi/pdf/10.1111/mec.15197, and get a behind-the-scenes look at this work below.

Remi Matthey‐Doret uses his new program SimBit to study the effects of background selection (BGS) on FST.

What led to your interest in this topic / what was the motivation for this study? 
It all started with a paper by Cruickshank and Hahn (2014), in which they highlight a fear that background selection could be a confounding factor to local adaptation in FST outlier studies. Curious about this issue, Mike and I investigated the question further and quickly figured that many of these fears were based on misinterpretation of Charlesworth et al. (1997). Indeed, Charlesworth et al. (1997) demonstrated that background selection can cause FST peaks for extreme and unrealistic parameter sets only. They highlighted that their parameter choice was unrealistic as their goal was to find extreme effects, but this important limitation of their study was sadly often ignored by their readers. We therefore decided to perform simulations of background selection with realistic parameter choices.

What difficulties did you run into along the way? 
The main difficulty was technical. We tried to run these simulations with a number of popular simulation softwares but none of them were fast enough for our needs. We quickly realized that we had to write our own simulation software (SimBit) that would have a very high performance especially for simulations with a lot of genetic diversity. 

What is the biggest or most surprising finding from this study? 
Starting the study, I was actually expecting that background selection would have a stronger effect on FST and that it would bias FST outlier methods to detect local adaptation. Our finding was a surprise to us, but it was also comforting to realize that the results of the many studies using FST outlier methods were probably not affected by background selection. 

Moving forward, what are the next steps for this research? 
I think there is a need for a clarified view of the relative importance of positive and negative selection in explaining patterns of genetic diversity within and between populations. Also, I would wish to investigate further the interaction between selection coefficient and migration rate and how it affects within and between population genetic diversity. Such an endeavor would likely require a mixture of empirical and theoretical work.

What would your message be for students about to start their first research projects in this topic?  
I think there is a lot of intuition about the effect of linked selection in structured populations that has not been published. Talk to smart people! They may have some expectation about how background selection can affect the coalescent tree in structured populations that needs to be studied and written out.

What have you learned about science over the course of this project? 
I learned that a lot of the numeric tools that we use to analyse genetic data contain bugs (one of which is detailed in our article) and untold (or somewhat neglected) assumptions. One must always be very careful to have a good understanding about a particular statistical software before using it.

Describe the significance of this research for the general scientific community in one sentence.
We found that background selection does not cause peaks of population differentiation and therefore that methods that use population differentiation to detect positive selection should be safe to be used without worry of background selection being a confounding factor.

Describe the significance of this research for your scientific community in one sentence.
We found that background selection does not cause much variation in locus-to-locus variation in FST and therefore FST outlier methods to detect positive selection should be safe to be used without worry of background selection being a confounding factor.

Full article:

Matthey‐Doret R, Whitlock MC. Background selection and FST: Consequences for detecting local adaptation. Mol Ecol. 2019;28:3902–3914. https://doi.org/10.1111/mec.15197.