Complex Patterns of Admixture across the Indonesian Archipelago

Abstract

Indonesia, an island nation as large as continental Europe, hosts a sizeable proportion of global human diversity, yet remains surprisingly undercharacterized genetically. Here, we substantially expand on existing studies by reporting genome-scale data for nearly 500 individuals from 25 populations in Island Southeast Asia, New Guinea, and Oceania, notably including previously unsampled islands across the Indonesian archipelago. We use high-resolution analyses of haplotype diversity to reveal fine detail of regional admixture patterns, with a particular focus on the Holocene. We find that recent population history within Indonesia is complex, and that populations from the Philippines made important genetic contributions in the early phases of the Austronesian expansion. Different, but interrelated processes, acted in the east and west. The Austronesian migration took several centuries to spread across the eastern part of the archipelago, where genetic admixture postdates the archeological signal. As with the Neolithic expansion further east in Oceania and in Europe, genetic mixing with local inhabitants in eastern Indonesia lagged behind the arrival of farming populations. In contrast, western Indonesia has a more complicated admixture history shaped by interactions with mainland Asian and Austronesian newcomers, which for some populations occurred more than once. Another layer of complexity in the west was introduced by genetic contact with South Asia and strong demographic events in isolated local groups.

Keywords: genetic admixture; genetic diversity; human migration; population structure.

Fig. 1.

Sampling locations and overview of genomic diversity. (A) Locations of new and published data from Mainland and Island Southeast Asia, Melanesia, and Polynesia used in the present study. Colors indicate regional affiliation of populations: yellow: mainland Asia, blue: Taiwan and the Philippines, magenta: western Indonesia, green: eastern Indonesia, purple: Melanesia, and turquoise: Polynesia. Detailed sample information is given in supplementary table S1, Supplementary Material online. (B) Principal Components Analysis of genome-wide SNP diversity in 703 individuals from the 32 populations shown in panel A. Note the different percentage variances explained by the two principal components.

Fig. 2.

Ancestral genomic components in regional populations. For every K, the modal solution with the highest number of ADMIXTURE runs is shown; individual ancestry proportions were averaged across all runs from the same mode and the number of runs (out of 50) assigned to the presented solution is shown in parentheses. Average cross validation statistics were calculated across all runs from the same mode (insert). The minimum cross-validation score is observed at K = 9. Note major ancestry components in Indonesia and ISEA—Papuan (light purple), mainland Asian (light yellow), and AN (light blue)—as well as major differences in the distribution of these three ancestries between eastern and western Indonesia. Populations from the Philippines and Flores are abbreviated as “Ph.” and “Fl.,” respectively.

Fig. 3.

Inference of population structure from haplotype sharing. (A) Simplified coancestry heatmap generated by chromosome painting with fineSTRUCTURE. Donor populations are shown in columns, recipients are in rows. Each cell depicts the average proportion of the genome copied by samples in the recipient population from the respective donor group, weighted by the average number of chunks received from this donor. Each row is scaled between 0 (blue, the minimum proportion of the genome copied for the given recipient group) to 1 (red, the maximum proportion of the genome copied). As expected, the largest proportion of the genome is copied from the population’s own label. The diagonal (gray), representing self-copying, was excluded from the scaling procedure to emphasize copying patterns from other donor populations. Recipient groups from outside ISEA, Melanesia, and Polynesia are not shown. A detailed matrix of copying vectors is given in supplementary figure S2, Supplementary Material online. (B) Simplified dendrogram showing the clustering of individuals with similar copying vectors into genetic groups. These groups were subsequently used in the main GLOBETROTTER analysis. Clusters are shaded according to the color scheme from figure 1A. A detailed dendrogram is given in supplementary figure S1, Supplementary Material online; individual sample affiliations are given in supplementary table S1, Supplementary Material online.

Fig. 4.

Population admixture events and inferred contact dates as inferred by the main GT analysis. Individual raw FS populations were grouped into larger population clusters (see fineSTRUCTURE results, fig. 3B and supplementary fig. S1 and table S1, Supplementary Material online) for the GLOBETROTTER analysis. Inferred admixture events were classified into three types: one-date admixture (OD, involving a pair of mixing sources); one-date-multiway admixture (MW, involving two pairs of mixing sources); and uncertain (U). The inferred composition of mixing sources is shown as barplots (one pair for OD and two pairs for MW). A pair of “best-guess” surrogate populations that matches the inferred contributing sources for the primary admixture event are shown as shaded triangles. The bean plots in the center depict the distributions of admixture dates as estimated by 100 bootstrap replicates. Dates inferred by “full” and “regional” analysis are colored in red and blue, respectively. White circles represent point estimates of the date of admixture. The color scheme assigned to the additional mainland East Asian (CHB and CHS), Southeast Asian (CDX and KHV), South Asian (STU), and European (CEU) surrogates is shown at the bottom of the plot. Surrogates contributing <3% to mixing sources are hatched in grey. Detailed information about the inferred admixture events and composition of mixing sources is given in supplementary table S3, Supplementary Material online. (A) Results of the “full” analysis. Each target population was allowed to copy from all donor populations, excluding self-copying from its own group. (B) Results of the “regional” analysis. Populations were divided into five geographical regions: the Philippines, western Indonesia, eastern Indonesia, Melanesia, and Polynesia (fig. 1A and supplementary table S1, Supplementary Material online). Each target group was allowed to copy from all donor populations, excluding populations with the same regional label (supplementary table S4, Supplementary Material online).

Fig. 5.

Alternative genomic dating of admixture events with MALDER. Each target Indonesian population (vertical axis) was represented as a mixture of Papuan (PNG) sources, and either mainland Asian (CHB, CHS, CDX, and KHV) or AN (Philippine Kankanaey and non-Kankanaey) sources. Point estimates and the standard error of dates for statistically significant admixture events are shown on the horizontal axis.