TS-IBD: An efficient ancestral recombination graph-based identity by descent segment detection method.

The ancestral recombination graph (ARG) provides a comprehensive framework for representing the evolutionary history of genome sequences. Advances in ARG inference have enabled the extraction of informative, low-dimensional genomic features, such as identity by descent (IBD) segments, which are continuous genomic intervals remaining uninterrupted by recombination events. Extracting IBD segments requires explicit ARGs, and developing efficient algorithms for this purpose remains an active area of research. Here, we introduce TS-IBD, an efficient method that leverages the tree sequence formalism of ARGs to extract recombination-based IBD segments. TS-IBD is optimized for detecting short IBD segments and for use in memory-constrained environments. We show that IBD segments inferred by TS-IBD exhibit threefold lower inflation than those derived from genotype-based methods when analyzing IBD segment coverage in centromeric regions. Additionally, we compare our results with IBD segments inferred by alternative methods, showing that in simulated datasets with ground-truth ARGs, TS-IBD captures more recombination events in distant admixture than other methods. The TS-IBD program is available at https://github.com/ucfcbb/TS-IBD.