Dstatistic

D-statistic, also known as Patterson’s D-statistic or the ABBA-BABA test, is a statistical measure used in population genetics to test for admixture between populations. It analyzes genome-wide patterns of allele sharing across four taxa: two reference populations, a putative admixed population, and an outgroup. The method relies on biallelic sites and distinguishes ancestral and derived alleles to count ABBA and BABA pattern occurrences.

The statistic is computed from the counts of ABBA and BABA site patterns across the genome. Let nABBA denote sites showing the ABBA pattern and nBABA denote sites showing the BABA pattern. D is defined as (nABBA − nBABA) / (nABBA + nBABA). In allele-frequency implementations, the statistic can be weighted rather than simply counted. Under a simple, non-admixture tree, ABBA and BABA patterns occur with equal frequency due to incomplete lineage sorting, yielding D near zero. A significant deviation from zero suggests gene flow between the test population and one of the other populations in the quartet.

Significance is typically assessed using block jackknife or bootstrap approaches over large genomic blocks to estimate standard errors and Z-scores. A |Z| value above common thresholds (for example, around 3) indicates statistical support for admixture.

Applications of the D-statistic are widespread in population genetics and evolutionary biology. It has been used to detect Neanderthal and Denisovan gene flow into modern humans and to explore introgression in various species. Limitations include sensitivity to outgroup choice, incorrect tree topology, data quality, and ancestral population structure, and it does not quantify the proportion of admixture, only its presence or absence.