“…Based on the phylogenetic tree, in order to avoid having an impact on the multi-species tracing of ancestral patterns, one sample of Sinocyclocheilus grahami and four samples of Sinocyclocheilus tingi were censored here, which was then filtered with 'AC==0 || AC==AN || F_MISSING >0.2' and the distance between the locus and the set SNPs to be no less than 100 bp, and ultimately reassembled them (2,642,407 SNPs for 103 individuals). For time calibration, due to the absence of a specific fossil record provides a calibration for Sinocyclocheilus, we used the time inferred by Chen et al (2009) to the split between S. grahami (F) and S. grahami (M) at approximately 2.37~0.47 Ma (Chen et al, 2009), Mao, Liu, Vasconcellos, et al (2021) to the split between S. oxycephalus and S. qujingensis at approximately 4.0~2.0 Ma (Mao, Liu, Vasconcellos, et al, 2021) and Yang et al (2021) to the split between S. grahami and S. tingi at approximately 0.68 Ma (Yang et al, 2021), and set the differentiation rate of S. grahami at 3.51 × 10 −9 per year per nucleotide (Yang et al, 2016). We ran an independent SNAPP analysis with 20 million MCMC generations, sampling at every 2000 steps.…”