Demographic History Inference
Hybrid individuals were removed from downstream analyses due to the
levels of heterozygosity, and the remaining individuals were used for
the inference of demographic history. The gene flow between each lineage
was inferred based on the allele frequency by applying the ABBA-BABA or
D statistic using the qpDstat module in AdmixTools 7.0 (Patterson et
al., 2012). Based on the phylogeny (((P1, P2), P3), O), the four
topologies were tested, with Paraquilegia microphylla andSemiaquilegia adoxoides as outgroups. In addition to the D
statistic, we performed gene flow analysis between different populations
using allele frequency data with 1-10 migration events by TreeMix v1.1
(Pickrell & Pritchard, 2012). The Python script easySFS was used to
calculate the joint site frequency spectrum (SFS) for demographic
analysis (https://github.com/isaacovercast/easySFS). We also calculated
the likelihood of different demographic scenarios in fastsimcoal2
software (Excoffier et al., 2021) using the joint SFS to infer
demographic parameters. Twenty scenarios were set up involving genetic
structure and gene flow, including three monophyletic models and sixteen
paraphyletic models (Figure S8). For each scenario, fastsimcoal2
performed 10000 coalescent simulations to approximate the expected SFS
in each cycle and will run 40 optimization cycles to estimate the
parameters. To ensure the accuracy of evaluating the best scenario, each
scenario was run 100 times, and the run with the highest likelihood was
compared by calculating the Akaike information criterion (AIC) to
determine the best scenario. The parameter estimation was run under the
best scenario 100 times with each of bootstrapped SFS. A neutral
mutation rate of 10-8 and a generation time of 1 year
were used to estimate the effective population size, divergence times
and migration rates (M. Li et al., 2019).