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.