Pseudoreplication in genomic-scale data sets.

Area of Science:

• Population Genetics
• Genomics
• Statistical Genetics

Background:

• Genomic-scale datasets contain correlated loci, leading to pseudoreplication.
• Pseudoreplication reduces effective degrees of freedom (df") compared to nominal degrees of freedom (df).
• The impact of pseudoreplication on genomic data has not been systematically quantified.

Purpose of the Study:

• Quantify pseudoreplication (df") for F_ST and r² across the genome.
• Assess the relationship between df", population size (N_e), and genome size.
• Evaluate the impact of individual sampling versus gene sampling on effective degrees of freedom.

Main Methods:

• Simulated genomic data using SLiM and msprime.
• Estimated df' and df"/df by measuring variance decline of mean F_ST and mean r² with increasing loci.
• Utilized variance components analysis to identify limiting factors for df'.

Main Results:

• Effective degrees of freedom (df") increase with N_e and genome size for both F_ST and r².
• df" for r² plateaus due to individual sampling limitations, even in large datasets.
• Pseudoreplication is less severe for F_ST, but significant df"/df ratios (<0.01) can occur with tens of thousands of loci.
• Block-jackknife methods overestimated variance for F_ST, yielding conservative confidence intervals.

Conclusions:

• Pseudoreplication significantly impacts statistical power in genomic studies.
• Predictive models for df' based on N_e, L, S, and genome size offer robust precision quantification.
• Accurate estimation of effective degrees of freedom is crucial for interpreting genomic-scale data.