Table 1 Statistical tests of neutrality

Within-species tests

Site-frequency spectrum

Most thoroughly studied

Powerful only if non-neutral evolution occurred within a critical time period

Coalescent likelihood methods

In principle, more powerful than summary statistic methods, possible to estimate multiple parameters simultaneously

Computationally intensive

F_ST

Does not require sequence data, can be performed with marker genotypes only

Low power to detect balancing selection

Long-range haplotype test (LRH)

Does not require sequence data, can be performed with marker genotypes only

Sensitivity of LRH test to population demographics and haplotype construction not well studied; not applicable to detecting balancing selection

Within- and between-species tests

Hudson - Kreitman - Aguade (HKA)

May be useful for detecting balancing selection

Requires sequences from multiple individuals in two species; may be difficult to interpret significant HKA test

McDonald - Kreitman

More sensitive than raw measure of the ratio of one number of non-synonymous amino acid substitution in a gene to the number of synonymous substitutions (d_n/d_s); may be fairly robust to population demographics and recombination

Requires sequences from multiple individuals in two species; selection to change codons may adversely affect test; applicable to protein-coding regions only

Between-species tests

Ratio of non-synonymous to synonymous substitutions

Requires only a single sequence from each species; raw d_n/d_s measure is unlikely to be confounded by demographics or recombination

Raw d_n/d_s measure is extremely stringent; applicable to protein-coding regions only