Genomic Effects of Nucleotide Substitutions in Drosophila simulans

Corresponding author: Andrew D. Kern, University of California, 1 Shields Ave., Davis, CA 95616., adkern{at}ucdavis.edu (E-mail)

Communicating editor: S. W. SCHAEFFER

	ABSTRACT

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Selective fixation of beneficial mutations reduces levels of linked, neutral variation. The magnitude of this "hitchhiking effect" is determined by the strength of selection and the recombination rate between selected and neutral sites. Thus, depending on the values of these parameters and the frequency with which directional selection occurs, the genomic scale over which directional selection reduces levels of linked variation may vary widely. Here we present a permutation-based analysis of nucleotide polymorphisms and fixations in Drosophila simulans. We show evidence of pervasive small-scale hitchhiking effects in this lineage. Furthermore, our results reveal that different types of fixations are associated with different levels of linked variation.

FIXATION of beneficial mutations results in reductions of linked, neutral variation. The scale of this hitchhiking effect (MAYNARD-SMITH and HAIGH 1974 ) reflects the relative strengths of selection and recombination during a selected mutant's sojourn through the population (HILL and ROBERTSON 1966 ; KAPLAN et al. 1989 ). In general, a reduction of linked variation is expected to occur over a physical region defined by the ratio of the recombination rate per base pair per generation and the selection coefficient. Therefore, if many beneficial nucleotide fixations result from weak selection, the associated hitchhiking effects may occur over only tens or hundreds of bases (KAPLAN et al. 1989 ). Despite recent work demonstrating that regional levels of nucleotide polymorphism across the genome in Drosophila and other organisms are affected by selection at linked sites (BERRY et al. 1991 ; BEGUN and AQUADRO 1992 ; LANGLEY et al. 1993 ; AQUADRO et al. 1994 ; NACHMAN 1997 , NACHMAN 2001 ; NACHMAN et al. 1998 ), the possibility of small-scale selective perturbations of polymorphism has not been investigated.

We can study the importance of small-scale hitchhiking effects by determining if levels of polymorphism are reduced near nucleotide sites that have fixed in the recent past. If some fraction of such sites fixed under directional selection, we may observe less heterozygosity in regions flanking fixations compared to randomly selected regions of DNA. Moreover, a priori categorization of fixations allows us to investigate possible heterogeneity of the substitution process across mutant classes. For example, if a greater fraction of amino acid fixations result from selection (compared to silent or noncoding fixations), the level of polymorphism in regions near amino acid fixations may be reduced relative to that in regions near silent fixations. More generally, sites experiencing stronger or more recent directional selection should be associated with regions of lower heterozygosity.

Here we develop a permutation-based test for detecting small-scale reductions of heterozygosity. Using standard methods from meta-analysis (FISHER 1935 , FISHER 1954 ; GLASS 1976 ; SOKAL and ROHLF 1995 ; GOOD 2000 ), we applied our permutation test to a sample of Drosophila simulans genes to see if particular types of fixations are associated with small-scale regions of reduced DNA polymorphism.

	MATERIALS AND METHODS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Sequence data:
The names, physical locations, and summary statistics of variation for the loci used in our analysis are in Table 1. Most of the sequence data we used are from BEGUN and WHITLEY 2000 and references therein (accession nos. AF204277–AF204290, AF256057–AF256078, AF252637–AF252824, AF255311–AF255314, AF255316–AF255320, AF255322–AF255327, AF255329, and AF256079). We did not use genes from regions of low recombination because our power to detect local reductions of polymorphism depends on the presence of several polymorphic sites per locus. Sequences for crq were provided by T. A. Schlenke (accession nos. AF544231 and AF544232–AF544239). Sequence data from Est-6 and Adh were obtained from GenBank [Est-6—(D. simulans) L34263, L34265, L10670, L34264, (D. melanogaster) AF147102, and (D. yakuba) AJ279007; Adh—(D. simulans) X57361, X57362, X57363, X57364, M36581, M19263, X00607, (D. melanogaster) X60792, and (D. yakuba) X57365].

All D. simulans-specific fixations were identified by parsimony using D. melanogaster and D. yakuba outgroup data. Parsimony should reliably identify ancestral states given the relatively low levels of sequence divergence between these species (YANG et al. 1995 ). Moreover, misidentified fixations should simply add noise across all of our analyses. To be conservative, we restricted our analyses to fixations having a single, most parsimonious reconstruction of the ancestral state. Synonymous codons were assigned to preferred vs. unpreferred classes according to SHARP and LLOYD 1993 . Both and (WATTERSON 1975 ; NEI 1987 ) were used as estimators of heterozygosity. Table 2 presents mean levels of heterozygosity for genes included in our analysis that contain at least one of the indicated class of fixation. Note that there does not seem to be a relationship between fixation class and polymorphism at the whole gene level.

View this table: In this window In a new window   Table 2. Mean and across classes of fixations

Permutation analysis:
The goal of the permutation analysis was to use DNA polymorphism data to empirically generate null distributions of nucleotide heterozygosities for windows of defined size. The null hypothesis is that levels of DNA polymorphism in regions near sites that fixed in a gene along the D. simulans lineage ("test sites") are the same as levels of polymorphism observed in randomly selected regions within the same gene. The alternative hypothesis is that regions near test sites have reduced heterozygosity. Each test site was assigned to a category of fixation (i.e., unpreferred, preferred, replacement, silent). For each gene, we estimated and for a window of 200 bp centered on each test site of a given category and then calculated the mean heterozygosity across test sites. Test site windows could overlap if fixations were close to one another. We then permuted the locations of test sites within each gene by centering windows of the same size on "randomly" selected sites (permuting locations of test sites rather than polymorphic sites preserves any underlying physical heterogeneity of polymorphic sites in a gene). Given that third positions of codons tend to be more variable than first and second positions, randomly selected sites in a gene may not reflect the underlying distribution of test sites across codon positions in real data. Thus, not all sites are exchangeable (GOOD 2000 ). To address this statistical problem, the fraction of first, second, and third positions sampled during the permutation step was made to equal the fraction of first, second, and third positions found at test sites. The number of sites selected was equal to the number of test sites in the actual data. As was the case for test sites, windows of random sites could overlap. Average and were then estimated for these randomly selected windows. This process was repeated several hundred times for each gene to generate a distribution of heterozygosities for windows of a given size. The total number of permutations depended on the number of exchangeable sites, which is approximately equal to the number of codons in the sample.

The observed mean heterozygosity across test sites for each gene was compared to the permutation-generated distribution. This yields the probability that the observed and for windows around test sites were lower than expected under the null hypothesis (thus, this is a one-tailed test). A potential problem with our approach could be that the ends of surveyed DNA sequences were undersampled during the permutations as windows exceeding the ends of the sequences were excluded from the analysis. However, this is not a major concern as heterozygosities at the 5' and 3' ends of the surveyed regions were not significantly different from heterozygosities at other regions (analysis not shown).

The choice of window size is a complex and important issue that may affect the picture of variation within a gene and the power of our analysis (SILVERMAN 1986 ). If window sizes are too small, then most randomly selected windows will have no segregating sites. This would reduce our power to detect a significant association of test sites with regions of reduced heterozygosity. Alternatively, large windows may "smooth" interesting local variation in polymorphism if the extent of the hitchhiking effect was substantially smaller than that of the window size (similar to the case of a large window width parameter in SILVERMAN 1986 ). A priori knowledge of the effect of window size on the power of rejecting the null hypothesis vs. some alternative for each gene would be ideal. In the absence of such information, however, we chose the window size on the basis of the following empirically based rationale. Although D. simulans is a relatively highly polymorphic species, most sites are monomorphic. The mean number of base pairs between polymorphic sites in our samples is 75 bp, with a standard deviation of 25 bp. A 200-bp window is approximately twice the mean number of base pairs between polymorphic sites (75 bp) plus one standard deviation (25 bp). This means that most randomly chosen windows of 200 bp will include at least one polymorphic site. We used the mean plus a standard deviation because using only the mean would have reduced our power in genes with less polymorphism, whereas increasing the window size was likely not biased. Regardless, the results using a window size based on the mean were not significantly different from those using a window size based on the mean plus a standard deviation (data not shown). Software and source code implementing this method are available from the authors and at http://limulus.ucdavis.edu/~cojo/.

Statistical analysis:
Failure to reject the null hypothesis at a locus may reflect a lack of polymorphism at that locus or other factors limiting the power of our analysis. We used Fisher's combined probability test to effectively increase our ability to detect a significant trend in the data. This test is suitable when separate statistical tests on different data sets test the same scientific hypothesis (FISHER 1954 ; SOKAL and ROHLF 1995 ) and is especially applicable when the same significance tests are used on all data sets, yet a joint statistical analysis is impossible (as is the case here). Moreover, our use of Fisher's combined probability test is likely to be unbiased because all loci were subjected to an identical analysis that yields exact probability values (SOKAL and ROHLF 1995 ).

One of the limitations of Fisher's test is that it cannot distinguish between several tests with consistent weak effects vs. a mixture of tests with strong effects and tests with no effect. To address this limitation, we used a standard test statistic from metaanalysis, Glass's g:

(GLASS 1976 ). Here, ^E is the mean level of polymorphism in windows surrounding test sites, ^C is the mean level of polymorphism in windows surrounding all sites of the gene, and s^c is the standard deviation of polymorphism in windows surrounding all sites of the gene. Thus, g is a unitless measure of the reduction (or inflation) of polymorphism surrounding test sites in a gene.

Using g has two main advantages. First, it allows us to estimate the relative magnitude of the reduction in heterozygosity near test sites from different categories of fixation. Second, g scores can be used to compare the relative reduction of polymorphism surrounding test sites across loci. If, on average, there is no difference in the levels of heterozygosity surrounding test sites and the levels of heterozygosity at all sites then the mean g across loci should be zero. A t-test can be used to test the null hypothesis that g is zero (i.e., there is no effect). This is an improvement over Fisher's test in that the null is less likely to be rejected if there are only a few genes of strong effect and many genes of no effect.

	RESULTS

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Table 3 and Table 4 show the results of permutation analyses of polymorphism in 200-bp windows centered on replacement and silent fixations, respectively. Heterozygosity near replacement fixations ( = 0.0067, = 0.0073) is slightly, though not significantly, reduced compared to overall levels of heterozygosity ( = 0.0087, = 0.0094) in sequenced regions of individual genes (Fisher's combined probability, d.f. = 16, P = 0.15 and P = 0.08 for and , respectively). To avoid the confounding effects of pooling data across loci, we calculated Glass's g statistic for each locus [g is a dimensionless measure of the difference between heterozygosity at our test sites and that of the gene as a whole (GLASS 1976 )]. An average g of 0 is expected under neutrality, whereas a negative g is expected if heterozygosity has been reduced. Consistent with the statistical results of average heterozygosity, mean g, while negative, is not statistically different from 0 (g = -0.42, t = -1.502, P = 0.15 for ; g = -0.379, t = -1.195, P = 0.25 for ). Similarly, heterozygosity in 200-bp windows centered on silent fixations ( = 0.0077; = 0.0081) is not significantly reduced compared to overall levels of heterozygosity ( = 0.0084, = 0.0086) in sequenced regions of individual genes (Fisher's combined probability, d.f. = 50, P = 0.06 and P = 0.054 for and , respectively). Interestingly, mean g across all loci for silent fixations is significant for (g = -0.484, t = -2.726, P = 0.0118) and is marginally significant for (g = -0.385, t = -2.049, P = 0.0516). None of the individual genes shows significantly reduced heterozygosity near replacement or silent fixations when critical values are Bonferroni corrected for multiple tests.

Genomic patterns of codon usage in Drosophila suggest that silent mutations can be placed into at least two categories, preferred and unpreferred (SHARP and LLOYD 1993 ). Therefore, we can investigate whether patterns of linked variation differ between mutant classes at silent sites. Mean heterozygosity and mean g (Table 5) reveal no reduction of heterozygosity for windows centered on unpreferred fixations ( = 0.0079, = 0.0086). Furthermore, no individual genes showed a significant reduction of polymorphism in windows centered on unpreferred fixations. Windows centered on preferred fixations, however, show a highly significant reduction of polymorphism (Table 6; d.f. = 46, P = 0.0096 and P = 0.0041 for and , respectively). Average polymorphism in regions near preferred codon fixations ( = 0.0065, = 0.0065) is 25% lower compared to the genes from which they were sampled ( = 0.0084, = 0.0086). This effect is confirmed by the mean g across all loci with preferred fixations (g = -0.612, t = -2.851, P = 0.0093 for ; g = -0.631, t = -3.159, P = 0.0045 for ), which also indicates a significant reduction in levels of heterozygosity flanking preferred fixations. Fig 1 provides a visual comparison of the distributions of g across loci for preferred and unpreferred fixations. Although five loci (AP-50, Cen190, crq, mei-218, and Pgd) show large reductions of for windows centered on preferred fixations, none are individually significant when critical values are conservatively adjusted for multiple tests.

View larger version (28K): In this window In a new window Download PPT slide   Figure 1. Comparison of loss of heterozygosity at preferred vs. unpreferred fixations. This histogram shows values of Glass's g for preferred (solid bars) and unpreferred (shaded bars) fixations at the per gene level. The distribution for preferred fixations is shifted toward negative values relative to the unpreferred distribution, indicating a reduction in local polymorphism and implicating the action of selection.

Fig 2 illustrates the effect of window size on our analysis. Fig 2A shows the results of an expanding window analysis for a strongly significant result, in this case a significant reduction around preferred sites. Clearly, the reduction in heterozygosity is statistically detectable for a variety of window sizes. Fig 2B and Fig C, shows results typical of nonsignificant genes. Both genes lack the long stretch of significant window sizes seen in Fig 2A. In Fig 2C, preferred fixations drop below P = 0.10 for windows of 140 bases, but only briefly, which suggests that this dip was due to chance.

View larger version (28K): In this window In a new window Download PPT slide   Figure 2. P values produced by permutation analysis for a series of window sizes. Window size began at 10 bases and increased by 10 bases until a width of 300 bases was reached or until the window around a test site extended beyond the ends of the sequence. Genes with all three classes of fixations [replacement (), unpreferred (), and preferred ()] were used in this analysis. (A) Results from mei-218, which showed a significant reduction in variation around preferred fixations in our earlier analysis and is intermediate in average heterozygosity. (B) Results from Osbp, which was not significant for any class of fixation and is intermediate in average heterozygosity. (C) Results from ry, which was not significant for any class of fixation and had the highest average heterozygosity of the 26 genes studied.

One concern regarding our analysis is that we have assumed that four-codon families can be represented as having only two fitness classes, preferred and unpreferred. However, conserved patterns of rank order of codon usage within codon families across widely divergent Drosophila (KREITMAN and ANTEZANA 2000 ) suggest that some four-codon families may have as many as four potential fitness classes. Although it is unclear under what circumstances this fact is problematic for our inference, we can avoid this complication by investigating silent fixations of mutants belonging to twofold codon families, which alternate between preferred and unpreferred states by reversible mutation. Congruent with the results from all preferred and unpreferred fixations, preferred and unpreferred classes in twofold codons are dramatically different in our estimate of their levels of linked heterozygosity (preferred, = 0.0055, = 0.0054; unpreferred, = 0.0091, = 0.0099), although this difference is not statistically significant (perhaps as the consequence of reduced power in this restricted data set).

	DISCUSSION

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

Our analysis of D. simulans polymorphism and divergence data revealed no evidence of hitchhiking effects associated with replacement fixations. One possible explanation is that the proteins in our sample evolve by genetic drift in D. simulans. Alternatively, replacement fixations may be composed of a large class of neutral mutants and a small class of strongly selected mutations. If this were the case we might not observe an overall association of replacement fixations with reduced heterozygosity. Finally, the power of our analyses of replacement fixations could be compromised if the physical scale of reduced variation near replacement fixations were greater than the size of the windows or gene regions used in our analyses. This explanation, however, seems unlikely because loci that have fixed at least one amino acid are roughly as polymorphic as those that have fixed only unpreferred mutations (Table 2).

We observed a reduction of linked polymorphism near preferred fixations, but not near unpreferred fixations (Table 7). One might expect this result under the simple premise that preferred and unpreferred mutations are slightly beneficial and slightly deleterious alleles, respectively. However, this expectation is probably incorrect because we are examining a special set of mutations, namely those that have fixed. MARUYAMA 1974 showed that conditional on fixation, the mean and variance of the sojourn time are the same for fixed beneficial and deleterious mutations having identical selection coefficients. Although counterintuitive, this result can be understood in the following way: given that a deleterious mutant will fix, the more strongly negative the mutant, the more quickly it must fix by drift to escape the selection that opposes its spread. Under this model, the hitchhiking effect associated with preferred and unpreferred fixations should be the same, all else being equal. Thus, one interpretation of our results is that some factor affecting the expected heterozygosity in a recurrent hitchhiking model differs among unpreferred and preferred fixations. For example, more recent fixations should be associated with greater reductions of linked polymorphism, all else being equal (KAPLAN et al. 1989 ; SIMONSEN et al. 1995 ; KIM and STEPHAN 2002 ; PRZEWORSKI 2002 ), because there has been less time for neutral variation to accumulate following the fixation event. Therefore, one hypothesis for the greater reduction of heterozygosity near preferred fixations is that, on average, preferred fixations have occurred more recently than other types of fixations.

Previous analyses suggested that the D. simulans lineage has fixed significantly more unpreferred mutations (BEGUN 2001 ; MCVEAN and VIEIRA 2001 ) than expected under the mutation-selection-drift model of silent-site evolution (BULMER 1991 ). An ancient accumulation of unpreferred mutations followed by a more recent, compensatory accumulation of preferred mutations in D. simulans is consistent with our data. What might cause such dynamics remains a matter for speculation. It is possible that D. simulans genes accumulated unpreferred fixations by drift when population size was small and subsequently fixed preferred codons by directional selection when population size increased. However, the timescale of changes in population size must be slower than that of the substitution rate for such a model to be viable. Alternatively, the intensity of selection favoring preferred mutations may vary over time. The potential impact of interactions among weakly selected polymorphisms on neutral variation (e.g., MCVEAN and CHARLESWORTH 2000 ; COMERON and KREITMAN 2002 ) also merits further consideration.

As is the case for silent sites, the lack of hitchhiking effects associated with amino acid fixations could be explained by invoking episodic evolution if these fixation events occurred in the more distant past compared to preferred fixations. This hypothesis may be testable if data from D. mauritiana and D. sechellia allow us to identify which mutations in the D. simulans lineage fixed in the more recent vs. more ancient past. It will also be interesting to investigate whether the spatial distribution of polymorphisms across D. melanogaster genes is similar to what we have observed in D. simulans.

Although we are not in a position to strongly favor a particular substitution model for our data, the results reported here certainly provide motivation for additional analyses of linked selection. For example, we have little understanding of how different population genetic parameters affect our permutation test or the population genetic scenarios under which we may be able to detect the local footprint of selection. Finally, our results underscore AKASHI's (1995, 1999) cautionary notes regarding the dangers of making population genetics inferences on the causes of protein evolution under the premise that silent mutations are neutral (e.g., SMITH and EYRE-WALKER 2002 ; FAY et al. 2002 ).

	FOOTNOTES

Sequence data from this article have been deposited with the EMBL/GenBank Data Libraries under accession nos. AF544231 and AF544232, AF544233, AF544234, AF544235, AF544236, AF544237, AF544238, AF544239.
² These authors contributed equally to this work.

	ACKNOWLEDGMENTS

We thank P. Awadalla, A. Betancourt, J. Gillespie, C. Langley, M. Lawniczak, M. Przeworski, S. Schaeffer, D. Weinreich, and two anonymous reviewers for comments on drafts of this manuscript. A.D.K. is a Howard Hughes Medical Institute predoctoral fellow. C.D.J. and D.J.B. were funded by the National Science Foundation.

Manuscript received April 22, 2002; Accepted for publication September 26, 2002.

	LITERATURE CITED

TOP ABSTRACT MATERIALS AND METHODS RESULTS DISCUSSION LITERATURE CITED

AKASHI, H., 1995  Inferring weak selection from patterns of polymorphism and divergence at silent sites in Drosophila. Genetics 139:1067-1076.[Abstract]

AKASHI, H., 1999  Inferring the fitness effects of DNA mutations from polymorphism and divergence data: statistical power to detect directional selection under stationarity and free recombination. Genetics 151:221-238.[Abstract/Free Full Text]

AQUADRO, C. F., D. J. BEGUN and E. C. KINDAHL, 1994 Selection, recombination and DNA polymorphism in Drosophila, pp. 46–56 in Non-neutral Evolution: Theories and Molecular Data, edited by B. GOLDING. Chapman & Hall, New York.

BEGUN, D. J., 2001  The frequency distribution of nucleotide variation in Drosophila simulans.. Mol. Biol. Evol. 18:1343-1352.[Abstract/Free Full Text]

BEGUN, D. J. and C. F. AQUADRO, 1992  Levels of naturally occurring DNA polymorphism correlate with recombination rates in D. melanogaster.. Nature 356:519-520.[Medline]

BEGUN, D. J. and P. WHITLEY, 2000  Reduced X-linked nucleotide polymorphism in Drosophila simulans.. Proc. Natl. Acad. Sci. USA 97:5960-5965.[Abstract/Free Full Text]

BERRY, A. J., J. W. AJIOKA, and M. KREITMAN, 1991  Lack of polymorphism on the Drosophila fourth chromosome resulting from selection. Genetics 129:1111-1117.[Abstract]

BULMER, M. G., 1991  The selection-mutation-drift theory of synonymous codon usage. Genetics 129:897-907.[Abstract]

COMERON, J. M. and M. KREITMAN, 2002  Population, evolutionary and genomic consequences of interference selection. Genetics 161:380-410.

FAY, J. C., G. J. WYCKOFF, and C.-I WU, 2002  Testing the neutral theory of molecular evolution with genomic data from Drosophila.. Nature 415:1024-1026.[Medline]

FISHER, R. A., 1935 The Design of Experiments. Hafner Press, New York.

FISHER, R. A., 1954 Statistical Methods for Research Workers, Ed. 12. Hafner Press, New York.

GLASS, G. V., 1976  Primary, secondary, and meta-analysis of research. Educ. Res. 5:3-8.

GOOD, P., 2000 Permutation Tests: A Practical Guide to Resampling Methods for Testing Hypotheses. Springer, New York.

HILL, W. G. and A. ROBERTSON, 1966  The effect of linkage on limits to artificial selection. Genet. Res. 8:269-294.[Medline]

KAPLAN, N. L., R. R. HUDSON, and C. H. LANGLEY, 1989  The "hitchhiking effect" revisited. Genetics 123:887-899.[Abstract/Free Full Text]

KIM, Y. and W. STEPHAN, 2002  Detecting a local signature of genetic hitchhiking along a recombining chromosome. Genetics 160:765-777.[Abstract/Free Full Text]

KREITMAN, M., and M. ANTEZANA, 2000 The population and evolutionary genetics of codon bias, pp. 82–101 in Evolutionary Genetics: From Molecules to Morphology, edited by R. S. SINGH and C. B. KRIMBAS. Cambridge University Press, Cambridge, UK.

LANGLEY, C. H., J. MACDONALD, N. MIYASHITA, and M. AGUADE, 1993  Lack of correlation between interspecific divergence and intraspecific polymorphism at the suppressor of forked region in Drosophila melanogaster and Drosophila simulans.. Proc. Natl. Acad. Sci. USA 90:1800-1803.[Abstract/Free Full Text]

MARUYAMA, T., 1974  The age of an allele in a finite population. Genet. Res. 23:137-143.[Medline]

MAYNARD SMITH, J. and J. HAIGH, 1974  The hitch-hiking effect of a favourable gene. Genet. Res. 23:23-35.[Medline]

MCVEAN, G. A. and B. CHARLESWORTH, 2000  The effects of Hill-Robertson interference between weakly selected mutations on patterns of molecular evolution and variation. Genetics 155:929-944.[Abstract/Free Full Text]

MCVEAN, G. A. and J. VIEIRA, 2001  Inferring parameters of mutation, selection and demography from patterns of synonymous site evolution in Drosophila. Genetics. 157:245-257.[Abstract/Free Full Text]

NACHMAN, M. W., 1997  Patterns of DNA variability at X-linked loci in Mus domesticus.. Genetics 147:1303-1316.[Abstract]

NACHMAN, M. W., 2001  Single nucleotide polymorphisms and recombination rate in humans. Trends Genet. 17:481-485.[Medline]

NACHMAN, M. W., V. L. BAUER, S. L. CROWELL, and C. F. AQUADRO, 1998  DNA variability and recombination rates at X-linked loci in humans. Genetics 150:1133-1141.[Abstract/Free Full Text]

NEI, M., 1987 Molecular Evolutionary Genetics. Columbia University Press, New York.

PRZEWORSKI, M., 2002  The signature of positive selection at randomly chosen loci. Genetics 160:1179-1189.[Abstract/Free Full Text]

SHARP, P. M., and A. T. LLOYD, 1993 Codon usage, pp. 378–397 in An Atlas of Drosophila Genes: Sequences and Molecular Features, edited by G. MARONI. Oxford University Press, Oxford.

SILVERMAN, B. W., 1986 Density Estimation for Statistics and Data Analysis. Chapman & Hall, London.

SIMONSEN, K. L., G. A. CHURCHILL, and C. F. AQUADRO, 1995  Properties of statistical tests of neutrality for DNA polymorphism data. Genetics 141:413-429.[Abstract]

SMITH, N. G. C. and A. EYRE-WALKER, 2002  Adaptive protein evolution in Drosophila.. Nature 415:1021-1024.

SOKAL, R. R., and F. J. ROHLF, 1995 Biometry, Ed. 3. W. H. Freeman, New York.

WATTERSON, G. A., 1975  On the number of segregating sites in genetical models without recombination. Theor. Popul. Biol. 7:256-276.[Medline]

YANG, Z., S. KUMAR, and M. NEI, 1995  A new method of inference of ancestral nucleotide and amino acid sequences. Genetics 141:1641-1650.[Abstract]

This article has been cited by other articles:

				A. Sanchez-Gracia and J. Rozas Unusual Pattern of Nucleotide Sequence Variation at the OS-E and OS-F Genomic Regions of Drosophila simulans Genetics, April 1, 2007; 175(4): 1923 - 1935. [Abstract] [Full Text] [PDF]

				R. Nielsen, V. L. Bauer DuMont, M. J. Hubisz, and C. F. Aquadro Maximum Likelihood Estimation of Ancestral Codon Usage Bias Parameters in Drosophila Mol. Biol. Evol., January 1, 2007; 24(1): 228 - 235. [Abstract] [Full Text] [PDF]

				C. D. Jones, A. W. Custer, and D. J. Begun Origin and Evolution of a Chimeric Fusion Gene in Drosophila subobscura, D. madeirensis and D. guanche Genetics, May 1, 2005; 170(1): 207 - 219. [Abstract] [Full Text] [PDF]

				J. Hermisson and P. S. Pennings Soft Sweeps: Molecular Population Genetics of Adaptation From Standing Genetic Variation Genetics, April 1, 2005; 169(4): 2335 - 2352. [Abstract] [Full Text] [PDF]

				A. D. Kern, C. D. Jones, and D. J. Begun Molecular Population Genetics of Male Accessory Gland Proteins in the Drosophila simulans Complex Genetics, June 1, 2004; 167(2): 725 - 735. [Abstract] [Full Text] [PDF]

				V. B. DuMont, J. C. Fay, P. P. Calabrese, and C. F. Aquadro DNA Variability and Divergence at the Notch Locus in Drosophila melanogaster and D. simulans: A Case of Accelerated Synonymous Site Divergence Genetics, May 1, 2004; 167(1): 171 - 185. [Abstract] [Full Text] [PDF]