Scalable bias-corrected linkage disequilibrium estimation under genotype uncertainty

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Linkage disequilibrium (LD) is crucial for genomic analyses like SNP pruning and decay estimation.
• Genotype uncertainty causes significant bias in naive LD estimates, particularly in polyploid organisms.
• Current accurate methods (maximum likelihood) are too slow for genome-wide applications.

Purpose of the Study:

• To develop scalable and accurate methods for estimating linkage disequilibrium (LD) in the presence of genotype uncertainty.
• To provide bias-corrected LD estimation suitable for genome-wide analyses, especially in polyploids.
• To introduce a computationally efficient alternative to existing methods.

Main Methods:

• Developed scalable moment-based adjustments to LD estimates using marginal posterior genotype distributions.
• Evaluated methods on both simulated and real genetic datasets.
• Implemented the methods in the R package 'ldsep'.

Main Results:

• Moment-based LD estimators demonstrate accuracy comparable to maximum likelihood estimation.
• These new estimators are significantly faster than maximum likelihood, approaching the speed of naive methods.
• The approach effectively reduces attenuation bias caused by genotype uncertainty.

Conclusions:

• Scalable, bias-corrected LD estimation is now feasible for genome-wide applications.
• The moment-based method provides a computationally efficient and accurate solution for LD estimation with genotype uncertainty.
• The 'ldsep' package offers a valuable tool for researchers in population genetics and genomics.