1 February 2014 in Diversity and Distributions
31 October 2019 in bioRxiv
Despite the increasing use of high-throughput sequencing in phylogenetics, many phylogenetic relationships remain difficult to resolve because of conflict between gene trees and species trees. Selection of different types of markers (i.e. protein-coding exons, non-coding introns, ultra-conserved elements) is becoming important to alleviate these phylogenomic challenges. For evolutionary studies in frogs, we introduce the new publicly available FrogCap suite of genomic resources, which is a large and flexible collection of probes corresponding to ∼15,000 markers that unifies previous frog sequencing work. FrogCap is designed to be modular, such that subsets of markers can be selected based on the phylogenetic scale of the intended study. FrogCap uses a variety of molecular marker types that include newly obtained exons and introns, previously sequenced UCEs, and Sanger-sequencing markers, which span a range of alignment lengths (100–12,000 base pairs). We tested three probe sets from FrogCap using 105 samples across five phylogenetic scales, comparing probes designed using a consensus- or genome-based approach. We also tested the effects of using different bait kit sizes on depth of coverage and missing data. We found that larger bait kits did not result in lowered depth of coverage or increased missing data. We also found that sensitivity, specificity, and missing data are not related to genetic distance in the consensus-based probe design, suggesting that this approach has greater success and overcomes a major hurdle in probe design. We observed sequence capture success (in terms of missing data, quantity of sequence data, recovered marker length, and number of informative sites) and compared them at all phylogenetic scales. The incorporation of different molecular marker types allowed recovery of the variation required for resolving difficult phylogenetic relationships and for performing population genetic studies. Altogether, FrogCap is a valuable and adaptable resource for performing high-throughput sequencing projects across variable timescales.
17 January 2025 in bioRxiv
Supervised machine learning approaches have proven powerful in population genetics. To use such approaches, training data with known inputs and outputs are required. Since such data are generally unavailable in population genetics, researchers typically rely on simulations under the models of interest to train machine learning algorithms. While powerful, this approach depends heavily on the models used to generate training data. Because of the variety and complexity of processes shaping genetic variation, it is inevitable that not all processes important in an empirical system will be included when generating training data. This leads to a mismatch between the data used to train a machine learning algorithm and the data to which the trained model is ultimately applied–i.e., a domain shift– and can negatively impact inference. Here, we train a Convolutional Neural Network (CNN) to detect introgression between sister populations and demonstrate that it has near perfect accuracy when applied to data generated under the models used for training. To evaluate the impacts of domain shifts on inference, we generated new data with introgression from a third, unsampled population into one of the two focal populations (i.e., ghost introgression), and accuracy was substantially reduced on these data. Finally, we used domain adaptation, which aims to train a network that performs well in the presence of a domain shift. Notably, this requires no knowledge of the target or empirical domain. Our domain adaptation network was able to accurately detect introgression, even in the presence of unmodelled ghost introgression. We also applied this approach to empirical data to detect introgression between ABC Island brown bears and other populations of brown bears. Previous work has suggested that introgression between ABC Island bears and polar bears can mislead tests of introgression between populations of brown bears. We found that using domain adaptation reduced support for introgression between geographically isolated populations of brown bears, suggesting that our approach reduces false inferences of introgression due to ghost introgression.