Ultraconserved Elements (UCEs) have been useful to resolve challenging phylogenies of non-model clades, unpuzzling long-conflicted relationships in key branches of the Tree of Life at both deep and shallow levels. UCEs are often reliably recovered from historical samples, unlocking a vast number of preserved natural history specimens for analysis. However, the extent to which sample age and preservation method impact UCE recovery as well as downstream inferences remains unclear. Furthermore, there is an ongoing debate on how to curate, filter, and properly analyze UCE data when locus recovery is uneven across sample age and quality. In the present study we address these questions with an empirical dataset composed of over 3800 UCE loci from 219 historical and modern samples of Sciuridae, a globally distributed and ecologically important family of rodents. We provide a genome-scale phylogeny of two squirrel subfamilies (Sciurillinae and Sciurinae: Sciurini) and investigate their placement within Sciuridae. For historical specimens, recovery of UCE loci and mean length per locus were inversely related to sample age; deeper sequencing improved the number of UCE loci recovered but not locus length. Most of our phylogenetic inferences—performed on six datasets with alternative data-filtering strategies, and using three distinct optimality criteria—resulted in distinct topologies. Datasets containing more loci (40% and 50% taxa representativeness matrices) yielded more concordant topologies and higher support values than strictly filtered datasets (60% matrices) particularly with IQ-Tree and SVDquartets, while filtering based on information content provided better topological resolution for inferences with the coalescent gene-tree based approach in ASTRAL-III. We resolved deep relationships in Sciuridae (including among the five currently recognized subfamilies) and relationships among the deepest branches of Sciurini, but conflicting relationships remain at both genus- and species-levels for the rapid Neotropical tree squirrel radiation. Our results suggest that phylogenomic consensus can be difficult and heavily influenced by the age of available samples and the filtering steps used to optimize dataset properties.

Human populations have been shaped by catastrophes that may have left long-lasting signatures in their genomes. One notable example is the second plague pandemic that entered Europe in ca. 1,347 CE and repeatedly returned for over 300 years, with typical village and town mortality estimated at 10%–40%.1 It is assumed that this high mortality affected the gene pools of these populations. First, local population crashes reduced genetic diversity. Second, a change in frequency is expected for sequence variants that may have affected survival or susceptibility to the etiologic agent (Yersinia pestis).2 Third, mass mortality might alter the local gene pools through its impact on subsequent migration patterns. We explored these factors using the Norwegian city of Trondheim as a model, by sequencing 54 genomes spanning three time periods: (1) prior to the plague striking Trondheim in 1,349 CE, (2) the 17th–19th century, and (3) the present. We find that the pandemic period shaped the gene pool by reducing long distance immigration, in particular from the British Isles, and inducing a bottleneck that reduced genetic diversity. Although we also observe an excess of large FST values at multiple loci in the genome, these are shaped by reference biases introduced by mapping our relatively low genome coverage degraded DNA to the reference genome. This implies that attempts to detect selection using ancient DNA (aDNA) datasets that vary by read length and depth of sequencing coverage may be particularly challenging until methods have been developed to account for the impact of differential reference bias on test statistics.