Data for: Beyond a single copy: Recovering paralogs from target capture datasets

Many commonly used target-capture probe sets are designed around putatively single-copy loci. However, large-scale gene family expansions (GFEs), including whole-genome and segmental duplications, are pervasive across angiosperms and can leave paralogous copies of targeted loci in lineages that have experienced duplications. Target-capture datasets often recover exons more reliably than the introns that connect them; assembly programs are not designed to differentiate the orthology of exon fragments when introns are disconnected. This can result in chimeric assemblies in which exons from different paralogs are combined into a single sequence, potentially generating spurious gene-tree conflict and misleading phylogenetic inference. We evaluate this in Caryophyllaceae, a lineage with a large-scale GFE, by simulating target-capture reads from reference genomes and assembling them using the popular assembly tool HybPiper. We show that unassembled connecting introns can lead to stitching of paralogous exon contigs. We then present ROOT, Reference-guided Ortholog Organization of Target-capture contigs, a workflow that uses phylogenetic information to group assembled contigs by orthologous copy and reduce paralog-derived chimerism. Using simulated and empirical target-capture data, we compare ROOT and HybPiper for recovering duplicated loci for downstream phylogenetic inference. ROOT provides a framework for improving target-capture assembly when targeted loci exhibit duplication histories.