Imputation in the Wild: Genome-Wide Robustness and Fine-Scale Limitations of Low-Coverage Genomes in Endangered...

Low-coverage whole-genome sequencing (lcWGS) combined with genotype imputation is increasingly being used to generate large genomic datasets at reduced cost, offering a promising alternative for conservation genomics. Here, we use the Iberian lynx (Lynx pardinus), a species with extremely low genetic diversity, genetically differentiated populations, and recent admixture, as a case study to evaluate the performance of imputed lcWGS data for population genomic inferences. We performed a leave-one-out imputation approach with 50 individuals sequenced to high coverage to assess imputation accuracy under different reference panel compositions, sequencing depths, and imputation strategies. We then compared inbreeding, admixture, and demographic history inferences derived from high-coverage and imputed datasets under highly concordant imputation scenarios (r² > 0.98; NRD rate < 0.05). Imputation accuracy substantially improved when using a mixed-ancestry reference panel, a multi-target imputation setting, and an adjusted effective population size parameter, particularly at low sequencing coverages. Genome-wide estimates of inbreeding and ancestry were largely robust. In contrast, fine-scale local genomic patterns were more sensitive to lower levels of imputation accuracy: long runs of homozygosity and ancestry tracts were reliably recovered, whereas shorter segments, which reflect older demographic events, were more error-prone. Consistent with these patterns, recent demographic trajectories were accurately reconstructed, while older effective population size estimates showed greater uncertainty. Overall, our results highlight the importance of validating and optimizing imputation performance in species with complex demographic histories and support lcWGS followed by imputation as a reliable framework for the study of recent demographic events in endangered species.