*4.2. RAD Sequencing and Data Processing*

RAD libraries were prepared using the restriction enzyme *Eco*RI and sample-specific barcodes. Samples in the libraries were pooled and sequenced on an Illumina HiSeq 2500 to generate 150 bp pair-end reads, at the Novogene Bioinformatics Institute (Beijing, China). Standard quality control (QC) pipelines were used to process the raw sequencing data. Reads with > 5% of unknown nucleotides, low-quality reads (quality value ≤ 5), and reads

that did not contain sample-specific barcodes were discarded. After filtering procedures, the quality-filtered reads data for each individual were then assembled into de novo loci, and SNP calling was performed using the following core programs: USTACKS, CSTACKS, SSTACKS, GSTACKS, and POPULATIONS, implemented in the STACKS pipeline v2.2 [41]. Briefly, all sequences were firstly processed in USTACKS to produce consensus sequences of RAD tags. The program USTACKS takes a set of short-read sequences as input and aligns them into exactly matching stacks. The minimum depth of coverage required for creating a stack was set at three (–m: 3) sequences, and the maximum distance allowed between stacks was set at four (–M: 4) nucleotides. The program CSTACKS was used to build a catalog of consensus loci containing all the loci from all the *H. chrysotricha* samples and merging all alleles together, with the number of mismatches allowed between sample loci set to four (–n: 4). Next, sets of stacks (i.e., putative loci) created in USTACKS were compared against the catalog using SSTACKS. Then, GSTACKS was used to align the reads to the locus and to call SNPs. Finally, the catalog of reads was filtered using the POPULATIONS program to produce a data set for downstream analyses. To identify highcredibility SNPs, polymorphic RAD loci that were present in at least 75% of the individuals in each population were retained (–r: 0.75). Potential homologs were excluded by removing loci showing heterozygosity of >0.5 within samples [42], and one SNP per RAD locus was kept, to reduce the impact of linkage disequilibrium.
