Geographical barrier identification and test of isolation by
distance
By calculating Monmonier’s maximum difference using the program BARRIER,
we inferred four strong geographic barriers. The robustness of each
barrier was assessed by bootstrapping over 100 matrices ofFST , with each matrix calculated from 30 genes
randomly sampled from the 93 sequenced genes. The first barrier,
supported by the 99 FST matrix, aligns with the
Malay Peninsula, isolating La-un and Nago from Chai-ya (Figure 5a). The
second barrier, with the support of 85 matrices, lies in the Indonesian
archipelago, isolated Bali from Kuching (Figure 5a). Both of the two
barriers are part of the Sundaland barrier. The third barrier lies in
the Wallacea region, supported with 93,75 and 75 matrices (Figure 5a).
This barrier isolated the populations in Australasia from those in
Southeast Asia. The fourth barrier, which is supported by 95, 95, and 59
matrices, lies between the northern South China Sea and southern South
China Sea (Figure 5a).
The Mantel test revealed a significant correlation (r =0.518,
P<0.01) between geographical distance and genetic
differentiation of Sanger FST estimations (Figure
5b). The same test using Solexa data also showed a significant
correlation (r=0.485, P-value < 0.01, Figure S7). The Mantel
tests within the “Pan-Australasia” group (r=0.560, P-value
<0.01, Figure S8) and “Indo-Malayan” group (r=0.698, P-value
<0.01, Figure S9) showed even more stronger correlations. This
indicated isolation by distance had contributed to the population
differentiation. Despite the significant r-value, the population
structures revealed before were evidenced by that the population pairs
within subgroups showed much smaller FST values
than the IBD expectation while the population pairs of s-SCS vs. n-SCS
or s-SCS vs. the Indian Ocean showed much largerFST (Figure 5b).