Abstract
The LDAandLDAS R package facilitates the computation of linkage disequilibrium of ancestry (LDA) and the linkage disequilibrium of ancestry score (LDAS). LDA quantifies the pairwise correlation of ancestral origins between single nucleotide polymorphisms (SNPs), while LDAS provides a cumulative measure of these correlations across genomic regions. These methods, originally described in Barrie et al. (2024), are essential for identifying genomic regions under "recombinant favoring selection," which has significant implications for understanding the evolutionary history of complex diseases.
Introduction
In the field of population genetics, traditional Linkage Disequilibrium (LD) measures the non-random association of alleles. However, in admixed populations, the ancestral origin of a genomic segment often provides more evolutionary insight than the specific DNA sequence itself. Quantifying how these ancestral assignments correlate across the genome has historically lacked accessible software. The LDAandLDAS package addresses this gap by providing a computational framework to measure these correlations and derive the LDA Score (LDAS), a metric used to detect regions where natural selection has favored novel ancestral combinations.
Methodology
The package implements algorithms designed to analyze "painting data"—the average ancestry probabilities for individuals across various SNPs. The core functionality revolves around two primary metrics:
- Linkage Disequilibrium of Ancestry (LDA): Calculated via functions like
cal_ldaandLDA, this metric measures the proportion of individuals who have experienced a recombination event that altered their ancestry at a specific locus relative to a genome-wide baseline. - LDA Score (LDAS): This function aggregates the total amount of the genome (in centiMorgans) that is in LDA with a specific SNP. A low LDAS serves as a signal for "recombinant favoring selection," indicating areas where the disruption of ancestral blocks was evolutionarily beneficial.
To handle the high computational cost of pairwise correlations across thousands of SNPs, the package utilizes Rcpp and supports parallel processing.
Data Structures and Implementation
The package includes built-in datasets to demonstrate its utility, such as example_map (genetic distances for 2,000 SNPs) and simulated painting data (example_painting_p1, example_painting_p2). Users typically input a list of data frames containing ancestry probabilities from different populations.
The output allows researchers to generate pairwise LDA matrices and map selection signals. By plotting LDAS against physical SNP positions, "dips" in the score can be visually identified, potentially highlighting regions associated with disease susceptibility or intense evolutionary pressure.
Limitations
While powerful, the package requires pre-processed ancestry inference data (e.g., from ChromoPainter). Additionally, while parallelization is supported, large-scale whole-genome analysis still demands significant computational resources. Currently, the high-speed runparallel feature for LDAS is restricted to Linux/Unix-based environments.
Summary updated