Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
, 361 (6401), 511-516

Evolutionary History and Adaptation of a Human Pygmy Population of Flores Island, Indonesia


Evolutionary History and Adaptation of a Human Pygmy Population of Flores Island, Indonesia

Serena Tucci et al. Science.


Flores Island, Indonesia, was inhabited by the small-bodied hominin species Homo floresiensis, which has an unknown evolutionary relationship to modern humans. This island is also home to an extant human pygmy population. Here we describe genome-scale single-nucleotide polymorphism data and whole-genome sequences from a contemporary human pygmy population living on Flores near the cave where H. floresiensis was found. The genomes of Flores pygmies reveal a complex history of admixture with Denisovans and Neanderthals but no evidence for gene flow with other archaic hominins. Modern individuals bear the signatures of recent positive selection encompassing the FADS (fatty acid desaturase) gene cluster, likely related to diet, and polygenic selection acting on standing variation that contributed to their short-stature phenotype. Thus, multiple independent instances of hominin insular dwarfism occurred on Flores.

Conflict of interest statement

Competing interests: J.M.A. is a paid consultant of Glenview Capital. R.E.G. is a paid consultant of Dovetail Genomics and Claret Biosciences.


Fig. 1.
Fig. 1.. Sampling location and genomic variation of the Flores pygmies.
(A) Location of Flores pygmy village and populations integrated into the analysis. The inset shows a subset of 85 populations from East Asia (EA), ISEA, and Oceania used for the PCA. The Rampasasa village (red square) is close to the Liang Bua cave, where H. floresiensis fossils were excavated. (B) PCA performed on 85 populations (13). (C) ADMIXTURE results for K = 6 clusters are shown for 96 worldwide populations (13).
Fig. 2.
Fig. 2.. Archaic hominin ancestry in the Flores pygmies.
(A) PCA to investigate genetic similarities of present-day humans and archaic hominins. Mean values for the top two principal components were plotted for each population. (B) Amounts of total archaic introgressed sequences in 9 Flores pygmies, 27 Melanesians, 103 East Asians, and 91 Europeans. The inset shows amounts of Neanderthal, Denisovan, and ambiguous sequences in Flores and Melanesian individuals. (C) The Denisovan D duplication is present only in Denisovan, Oceanic, and Flores individuals. A panel of worldwide populations is shown, along with the Denisovan and Neanderthal genomes (13). A copy number greater than two (four and three for light and dark blue, respectively) in region D (far right) indicates presence of the duplication.
Fig. 3.
Fig. 3.. Population genetic and functional signatures at the FADS locus.
(A) LocusZoom local Manhattan plot showing individual SNP PBS values spanning the FADS locus. Haplotype-defining SNP rs174547 is shown in dark gray, whereas other SNPs are colored according to pairwise linkage disequilibrium with rs174547 (from East Asian populations from the 1000 Genomes Project). (B) Geography of Genetic Variants map at rs174547. Data from Flores (n = 18 haplotypes), Melanesian (n = 54 haplotypes), and the Greenlandic Inuit (n = 4 haplotypes) populations are overlaid on populations from the 1000 Genomes Project. SNP array data from the Human Origin Dataset are shown in the inset. (C) Multitissue eQTL data from the GTEx (Genotype-Tissue Expression) Project, depicting associations between FADS1 and FADS2 expression and the rs174547 genotype. Effect size is displayed on the x axis and by color, whereas significance is indicated by point sizes.
Fig. 4.
Fig. 4.. Polygenic selection for reduced stature in the Flores pygmies.
(A) Comparison of the genome-wide genetic predictor of height from observed genotypes (purple) versus the null model (blue). The Flores panel is significantly enriched for height-reducing alleles (P < 0.001) in a multivariate chi-square test of population genetic differentiation from the expectation under the null model. (B) Frequency differences between the Flores population and neighboring 1000 Genomes Project populations for 4000 genome-wide loci of the strongest association, with height plotted against effect size for the height-increasing allele. The regression slope shown in the figure between the height-increasing effect size and the frequency difference was −0.71 (t = −3.18, P = 0.002), reflecting height-increasing alleles being at lower frequency in the Flores population. (C) Association between height in the Flores population and a genetic predictor of height.

Similar articles

See all similar articles

Cited by 4 PubMed Central articles

Publication types