Site frequency spectra from genomic SNP surveys

Area of Science:

• Population Genetics
• Evolutionary Genomics
• Bioinformatics

Background:

• Genomic survey data enable sensitive detection of deviations from evolutionary models, such as population size expansions and selective sweeps.
• Site frequency spectra (SFS) are crucial for inferring demographic history and selection from genetic variation data.

Purpose of the Study:

• To investigate how variations in sampling distributions affect goodness-of-fit analyses of SFS.
• To determine if different sampling strategies can mimic biological signals of selection or demographic changes.

Main Methods:

• Construction of site frequency spectra from single nucleotide polymorphism (SNP) data.
• Comparison of SFS generated under different sampling conditions, including conditioning on two alleles, a single mutational event, or random mutation selection.
• Goodness-of-fit analyses to assess departures from the standard biallelic model.

Main Results:

• Conditioning SFS on exactly two alleles renders them independent of the scaled neutral substitution rate (theta).
• Alternative sampling distributions (e.g., single mutational event, random mutation selection) yield distinct SFS.
• These distinct SFS show significant departures from the biallelic model predictions, potentially misinterpreting data filtering effects as biological signals.

Conclusions:

• Subtle differences in sampling distributions can profoundly influence SFS analyses and goodness-of-fit tests.
• Careful consideration of sampling strategies is essential to avoid spurious inferences of demographic events or selection.
• Data filtering procedures may inadvertently introduce biases that mimic violations of the standard neutral model.