Detecting past positive selection through ongoing negative selection

Abstract

Detecting positive selection is a challenging task. We propose a method for detecting past positive selection through ongoing negative selection, based on comparison of the parameters of intraspecies polymorphism at functionally important and selectively neutral sites where a nucleotide substitution of the same kind occurred recently. Reduced occurrence of recently replaced ancestral alleles at functionally important sites indicates that negative selection currently acts against these alleles and, therefore, that their replacements were driven by positive selection. Application of this method to the Drosophila melanogaster lineage shows that the fraction of adaptive amino acid replacements remained approximately 0.5 for a long time. In the Homo sapiens lineage, however, this fraction drops from approximately 0.5 before the Ponginae-Homininae divergence to approximately 0 after it. The proposed method is based on essentially the same data as the McDonald-Kreitman test but is free from some of its limitations, which may open new opportunities, especially when many genotypes within a species are known.

FIG. 1.—

Test for positive selection based on polymorphism at sites of ancestral divergence. (a) Change in the mode of selection as a result of an allele replacement. A fitness landscape that initially causes positive selection in favor of a rare allele “A” (left) causes negative selection against a rare allele “a” after “a”→”A” allele replacement is accomplished (right). Fitnesses are shown by vertical bars and allele frequencies are shown by pie charts. (b) Approach to measuring past positive selection. Extant species (dots) were used to infer allele replacements (“a”→”A”) that occurred at different segments (1−5; in the example shown, 3) of the ancestral lineage (see also supplementary fig. S1, Supplementary Material online). At sites of such replacements, the species for which polymorphism data is available (shown as a triangle) was used to assess the frequency of the ancestral variant (“a”). Such frequencies at nonsynonymous and synonymous sites were compared with measure β. (c–e) Results of the test for positive selection for substitutions which occurred in the lineage of the Drosophila melanogaster nuclear genome (c), Homo sapiens nuclear genome (d), or H. sapiens mitochondrial genome (e). The considered phylogeny is shown together with the times of the beginning and end of each segment of the lineage, measured in units of D_s from present. The species for which polymorphism data has been analyzed is shown as a triangle, with the number of available haploid genotypes N shown next to the species name. For each of the five considered segments, presented are the values of β together with 95% confidence intervals as horizontal bars, and the proportion of bootstrap replicates with β > 0 as pie charts.

FIG. 2.—

Performance of the MK test and the proposed test on mixtures of sites with different modes of selection. Orange solid line, the actual fraction of substitutions coming from the sites under switching selection; green triangles, α; yellow triangles, α with low-frequency (<15%) polymorphisms excluded; blue squares, β. (a–c) Horizontal axis: the fraction of switching sites (s = ±10⁻³); the remaining sites are neutral (a) or are under very weak (b, s = 10⁻⁵) or strong (c, s = 10⁻³) constant selection. (d) Horizontal axis: the fraction of sites under weak constant selection (s = 10⁻⁴); the remaining sites are neutral. (e–g) The fraction of switching sites (s = ±10⁻³) is 10% (e), 5% (f), or 3% (g); horizontal axis: the fraction of sites under weak constant selection (s = 10⁻⁴); the remaining sites are neutral.

FIG. 3.—

Distributions of the number of genotypes carrying the ancestral alleles at polymorphic sites. Only sites carrying the derived allele in >50% of the genotypes, and the ancestral allele in some of the genotypes, were taken into account. (a) Data for 6,726 human mitochondrial genotypes, for replacements that occurred in any of the five segments (Mann–Whitney U test for frequency of ancestral nonsynonymous vs. synonymous allele, n = 159, P = 0.053). (b) and (c) Data for 19 human nuclear genotypes, for replacements that occurred in segments 1–3 (b; n = 1184, P = 0.355) and 5 (c; n = 1870, P = 0.108); the excess of high-frequency polymorphism in this case is due to unfinished replacements. (d) Data for 37 Drosophila melanogaster genotypes, for replacements that occurred in segments 1–4 (n = 1892, P = 1.03 × 10⁻⁸). Nonsynonymous sites, purple; synonymous sites, blue. For the complete data set, see supplementary figures S2–S4 (Supplementary Material online).