Limits...
Evaluation of demographic history and neutral parameterization on the performance of FST outlier tests.

Lotterhos KE, Whitlock MC - Mol. Ecol. (2014)

Bottom Line: BayeScan was improved when the prior odds on neutrality was increased, regardless of the true odds in the data.In addition, default settings in FDIST2 and BayeScan resulted in many false positives suggesting balancing selection.We conclude that in species that exhibit IBD or have undergone range expansion, many of the published FST outliers based on FDIST2 and BayeScan are probably false positives, but FLK and Bayenv2 show great promise for accurately identifying loci under spatially divergent selection.

View Article: PubMed Central - PubMed

Affiliation: Department of Zoology, University of British Columbia, Vancouver, British Columbia, Canada, V6T 1Z4.

Show MeSH
Effect of prior odds on false-positive rates, power and false discovery rates (FDR) from BayeScan. Error rates were based on q-values (with a stated FDR of 0.01), while manipulating prior odds (PO) are based on empirical P-values (see Methods). False-positive rates based on empirical P-values were <0.002. Note the y-axis scale for false-positive rate is the same in Fig. 2, but larger than in Figs 5 and 6. (A–C) 9900 neutral loci and 100 selected loci; (D–F) 9000 neutral loci and 1000 selected loci. IM, island model; IBD, isolation by distance; 1R, expansion from one refuge; 2R, expansion from two refugia. Error bars are standard errors.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4228763&req=5

fig03: Effect of prior odds on false-positive rates, power and false discovery rates (FDR) from BayeScan. Error rates were based on q-values (with a stated FDR of 0.01), while manipulating prior odds (PO) are based on empirical P-values (see Methods). False-positive rates based on empirical P-values were <0.002. Note the y-axis scale for false-positive rate is the same in Fig. 2, but larger than in Figs 5 and 6. (A–C) 9900 neutral loci and 100 selected loci; (D–F) 9000 neutral loci and 1000 selected loci. IM, island model; IBD, isolation by distance; 1R, expansion from one refuge; 2R, expansion from two refugia. Error bars are standard errors.

Mentions: False-positive rates, power and false discovery rates (FDR) for the default settings in BayeScan (prior odds = 10), FDIST2 and FLK. Rates are based on q-values and a stated FDR of 0.01. Note the y-axis scale for false-positive rate is the same in Fig. 3, but larger than in Figs 5 and 6. (A,F) 10 000 neutral loci; (B,D,G) 9900 neutral loci and 100 selected loci; (C,E,H) 9000 neutral loci and 1000 selected loci. IM, island model; IBD, isolation by distance; 1R, expansion from one refuge; 2R, expansion from two refugia. Error bars are standard errors. False-positive rates for FDIST2 and FLK were zero in the island model, which led to a 0% FDR.


Evaluation of demographic history and neutral parameterization on the performance of FST outlier tests.

Lotterhos KE, Whitlock MC - Mol. Ecol. (2014)

Effect of prior odds on false-positive rates, power and false discovery rates (FDR) from BayeScan. Error rates were based on q-values (with a stated FDR of 0.01), while manipulating prior odds (PO) are based on empirical P-values (see Methods). False-positive rates based on empirical P-values were <0.002. Note the y-axis scale for false-positive rate is the same in Fig. 2, but larger than in Figs 5 and 6. (A–C) 9900 neutral loci and 100 selected loci; (D–F) 9000 neutral loci and 1000 selected loci. IM, island model; IBD, isolation by distance; 1R, expansion from one refuge; 2R, expansion from two refugia. Error bars are standard errors.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4228763&req=5

fig03: Effect of prior odds on false-positive rates, power and false discovery rates (FDR) from BayeScan. Error rates were based on q-values (with a stated FDR of 0.01), while manipulating prior odds (PO) are based on empirical P-values (see Methods). False-positive rates based on empirical P-values were <0.002. Note the y-axis scale for false-positive rate is the same in Fig. 2, but larger than in Figs 5 and 6. (A–C) 9900 neutral loci and 100 selected loci; (D–F) 9000 neutral loci and 1000 selected loci. IM, island model; IBD, isolation by distance; 1R, expansion from one refuge; 2R, expansion from two refugia. Error bars are standard errors.
Mentions: False-positive rates, power and false discovery rates (FDR) for the default settings in BayeScan (prior odds = 10), FDIST2 and FLK. Rates are based on q-values and a stated FDR of 0.01. Note the y-axis scale for false-positive rate is the same in Fig. 3, but larger than in Figs 5 and 6. (A,F) 10 000 neutral loci; (B,D,G) 9900 neutral loci and 100 selected loci; (C,E,H) 9000 neutral loci and 1000 selected loci. IM, island model; IBD, isolation by distance; 1R, expansion from one refuge; 2R, expansion from two refugia. Error bars are standard errors. False-positive rates for FDIST2 and FLK were zero in the island model, which led to a 0% FDR.

Bottom Line: BayeScan was improved when the prior odds on neutrality was increased, regardless of the true odds in the data.In addition, default settings in FDIST2 and BayeScan resulted in many false positives suggesting balancing selection.We conclude that in species that exhibit IBD or have undergone range expansion, many of the published FST outliers based on FDIST2 and BayeScan are probably false positives, but FLK and Bayenv2 show great promise for accurately identifying loci under spatially divergent selection.

View Article: PubMed Central - PubMed

Affiliation: Department of Zoology, University of British Columbia, Vancouver, British Columbia, Canada, V6T 1Z4.

Show MeSH