Fig 1.
The parameters of alternative selection models are depicted for the (A) 2013 and (B) 2014 data.
Hypothesis tests are expressed in terms of parameter constraints where p indicates reference base frequency: pA for reproductive adults, pM for successful male gametes, and pL plants for plants that fail to reproduce. H0 is the full neutral model. Male selection is tested by contrast of H1 to H0 in 2013 and H3 to H1 in 2014. Viability selection is tested by contrast of H3 to H2. (C) After DNA sequencing, read-pairs are mapped to the M. guttatus reference genome. The haplotype matching method (read-pairs to genic-haplotypes) is illustrated for a simple case with read-pairs mapping to single location. Read-pairs impose a probabilistic ‘process of elimination’ on reference line sequences as putative ancestors: √ indicates consistency and “X” inconsistency.
Fig 2.
The observed allele frequency change (2013 adults to 2014 zygotes) is compared to predicted with SNPs chosen based on (A) significance for male selection in 2013 (n = 587) or (C) significance for the Change in Allele Frequency test (n = 274). Results are reported for all gene sets with a SNP with p < 10−5. Contours indicate the density of points in panels A,C. For cross-validation (B, D), we split the data into two halves and performed model fits on each half. We chose an equivalent number of SNPs to the corresponding un-partitioned analyses with n = 587 in (B) to match (A) and n = 274 in (D) to match (C). For SNPs selected based on male selection (B), the “Ascertained” contrast is based on the predicted Δp from the significant test (orange points) while the “Paired” contrast is based on the predicted Δp from the other half of the data (blue points). (D) In the cross-validation for allele frequency change significant tests, the ascertained (orange) is the observed Δp from the significant test and predicted Δp from the other data half. Assignment is reversed because the allele frequency change test is based on the observed Δp.
Fig 3.
Testing haplotype matching: (A) In Mimulus, the precision of estimation is depicted as a function of the amount of data per plant. Compatible means that the likelihood for a genic-genotype is within 50% of the most likely genotype. (B) In Drosophila, the number of ancestors (indicated by contours and color) matching the genotype of a particular RIL is depicted as a function of amount of data (reads) and the number of SNPs in the gene set.
Fig 4.
Manhattan plots, with a single test reported per gene, for (a) Male selection 2013, (b) Allele frequency change 2013–2014, (c) Male selection 2014, and (d) Viability selection 2014. The orange line is the Bonferroni threshold, purple is p = 10−5.
Fig 5.
(A) Pairwise contrasts between predicted changes owing to male selection in 2013, viability selection in 2014, and male selection in 2014. A single SNP per gene is reported (the most significant) if p < 10−5. The SNPs ascertained for contrasts are distinct from those for tests in isolation (Fig 4), although these SNP sets are overlapping (see S5 and S9 Tables). (B) Density plots for molecular test statistics are depicted for 587 selected-SNP windows and 2751 control windows. Vertical lines are means. The number of polymorphisms (S), nucleotide diversity (pi), and Tajima’s D are based on the flanking DNA (100 bp on each side, but not including the focal SNP). Z focal is Zns calculated by contrasting the focal SNP to all flanking SNPs.