Restriction-site associated DNA sequencing (RADseq) is becoming popular in evolutionary and phylogenetic studies fostered by the relatively simple, fast and cost-effective data acquisition in non-model organisms. However, using this type of genomic data to resolve phylogenetic relationships among closely related species is hampered by missing data, difficulties in locus assembly and a low number of phylogenetically informative sites per locus. Here we test whether a modified ddRADseq protocol, which includes a pre- and a post-amplification size selection step to reduce the locus pool from tens of thousands to several thousand with a fixed length of 300-600bp, has the potential to generate a RAD dataset variable enough for resolving rapid radiations by incorporating coalescent-based summary methods in addition to widely used concatenated-based maximum likelihood approaches. Furthermore, we demonstrate how to assemble such long RAD fragments using the ipyrad pipeline and a customized clustering optimization approach, and give guidance for generating
filtered datasets for a comparative phylogenetic analysis with RAxML, ASTRAL III, SVDquartets and IQ-TREE. Using the plant genus Aichryson (Crassulaceae) as an example, we show that the long and variable loci dataset, generated by utilization of an optimized assembly/clustering and dataset filtering approach, supports maximum likelihood methods as well as coalescent-based summary methods for phylogenetic tree reconstructions. These are further evaluated with respect to gene- and site concordance factors using IQ-TREE, which in combination are able to reveal a complex phylogenetic history. We conclude that our llddRADseq approach is a powerful and efficient tool for phylogenomics especially in species-rich, non-model lineages. Less...