Predicting Genetic Relatedness from Low-Coverage Sequencing Data of Human and Animal Genomes Using Various Algorithms.

Background/

Objectives

The further application of high-coverage whole genome sequencing in fields such as paleogenomics, forensic investigations, and conservation genomics is impeded by two major barriers: extremely high costs and stringent sample requirements. Utilizing low-coverage sequencing offers a practical solution to these constraints; however, this approach introduces a primary challenge-the necessity to reconstruct distorted genomic information for downstream analysis.

Methods

Analytical experiments conducted on low- to medium-coverage sequencing data confirmed the accuracy of several existing methods for inferring relationships up to the third degree and distinguishing unrelated individuals. Subsequently, efforts were made to evaluate allele-frequency-independent methods within animal genomics, where analyses are likely to encounter challenges such as uncertain allele frequencies, diverse sample types, and suboptimal sample quality. Kinship inference was performed on a total of 33 pairs of animal samples across three species, comprising nine parent-offspring pairs and four full-sibling pairs.

Results

The analysis revealed that two efficient algorithm implementations (READ and KIN) successfully identified all unrelated pairs. Notably, among the various algorithms utilized, only KIN exhibited confusion between first- and second-degree relationships when subjected to.

Conclusions

This study has filled a critical gap in the existing literature by conducting a comprehensive evaluation of various algorithms on low-coverage sequencing data derived from authentic human and animal samples, accompanied by detailed ground truth-a vital task that has been overlooked.