USER MANUAL vs 2.

6: Software for the Measurement of Genetic Diversity (SMOGD)

1
1.1

Introduction
Purpose

SMOGD (Software for the Measurement of Genetic Diversity) is a web based application for the calculation of genetic diversity. Specically, it calculates GST est (Nei, 1983) GST (Hedrick, 2005) and Dest (Jost, 2008) diversity indices. It also generates bootstrap replicates of data sets and uses these replicates to estimate standard error, variance, and 95% condence intervals. 1.1.1 A Brief Note about Dest and GST
est

Although GST est , like Dest , better accounts for populations with high allelic diversity, only Dest accounts for populations alleles alternatively xed in dierent populations. To clarify: in a scenario where one subpopulation is xed at allele A, but ten other sub-populations are xed at allele B, GST est will over report a population-wide diversity of 1.0 while Dest will report a population diversity of 0.55.

1.2
1.2.1

Parameters Calculated
Basic Parameters: assumes actual allele frequencies are known

n = number of populations DST = absolute dierentiation (Nei, 1973) Equation: Ht Hs GST = relative dierentiation (Nei, 1973) Equation: Dst/Ht HST = between-subpopulation heterozygosity (Aczel & Daroczy, 1975; Tsallis & Brigatti, 2004) Equation:
(HtHs) (1Hs)

ST = between-subpopulation component of diversity, or the eective number of distinct subpopulations (Jost, 2008) Equation:
(1Hs)1 (1Ht)1

D = actual dierentiation (Jost, 2008)

 HS /HT = proportion intra-population heterozygosity vs total heterozygosity (Jost, 2008) S /T = proportion of total diversity that is contained in the average subpopulation (Jost, 2008) Equation: 1.2.2
(1Ht)1 (1Hs)1

Estimated Parameters: diversity measures for small sample sizes

= harmonic mean of population sizes N Equation:

approximate = approximation of harmonic mean as suggested by Anne Chao. It can handle neg N ative numbers and thus is use for calculating the harmonic mean of Dest values which can be negative. Equation: HS
est 1 (1/mean)+variance(D est )(1/mean)3

= nearly unbiased estimator of within-subpopulation heterozygosty (Nei & Chesser, 1983)

N ) Equation: ( (2(2 1) ) Hs N

HT

est

= nearly unbiased estimator of total-subpopulation heterozygosty (Nei & Chesser, 1983)

Equation: Ht + HST
est

= nearly unbiased estimator of between-subpopulation heterozygosity (Nei & Chesser, 1983)

Equation: GST
est

= nearly unbiased estimator of relative dierentiation (Nei & Chesser, 1983)

Equation: GST
est

= standardized measure of genetic dierentiation (Hedrick, 2005)

Equation:

Dest = estimator of actual dierentiation (Jost, 2008)

1.2.3

Bootstrap Parameters

Bootstrapped estimates of GST est , GST est , and Dest , (= values of diversity indices averaged across replicates). Variance and standard error of the mean calculated from bootstrap replicates using the standard equations from the Numpy python module. Generally it is not appropriate to report the boostrapped estimates of diversity. Rather, boostrapping is included to provide measures of condence. 1.2.4 Distance Matrices

Tables of pairwise distances for GST est , GST est , and Dest for each locus.

2
2.1

File Formats
Import

SMOGD will import les in the GenePop (Raymond & Rousset,1995) and Arlequin (Excoer et al., 1997). If your data is not in one of these formats I recommend using GenAlEx, an MS-Excel plugin for population genetic analysis to manipulate your data and export it in any one of the preceding formats. 2.1.1 GenePop

Details concerning GenePop format may be found at http://genepop.curtin.edu.au/ 2.1.2 Arlequin

Details concerning Arlequin format may be found at http://cmpg.unibe.ch/software/arlequin3/

2.2

Export

Data is exported as both html and tab-delimited les suitable for import into MS-Excel or database programs. The tab delimited les are time-stamped and the html links to these les are dynamically updated so that a user can only download les relating to the results of the data they submitted. Result les are deleted from the web-sever every 24 hours. Data sets are never saved although technically they exist in memory (RAM) until the user navigates away from the webpage.

3
3.1

Background
Basic and Estimated Parameters

SMOGD essentially calculates two sets of parameters: The basic parameters correspond to the diversity measures reported in Table 1 of Jost (2008). They are presented as illustrative examples of how the parameters dier from each other. For actual data sets, where you have genotypes of individuals sampled from larger populations, the estimated parameters more accurately account for small population sizes and associated sampling errors (Nei & Chesser, 1983). One common concern when interpreting the results is the presence of negative parameters. These occur when populations are almost identical/undierentiated. In these situations, the results can be reported as 0. However, SMOGD does not make this conversion for you.

3.2

Bootstrapping and Distance Matrices

Bootstrapping provides a way to estimate variance, standard error of the mean, and 95% condence intervals. Bootstrapping of subdivided population genetic data can be done at the population level (resampling populations) and the individual level (resampling individuals only) and at the individual and population level (resampling both populations and individuals). The implementation of the bootstrapping algorithm employed by SMOGD resamples at the individual level. Generally, bootstrapping to estimate parameters (e.g., averaging GST est or Dest across replicates) does not provide good measures of diversity (Petit & Pons, 1998). However, it can be used to estimate variance and standard deviation of the mean (Jost, 2008; Chao et al., 2008). The recommendation then is: dont report the bootstrap estimates of the estimated parameters, rather report the estimated parameters and the bootstrapped estimates of variance, standard deviation of the mean, and 95% condence intervals. I should also note that it is possible to calculate estimated values that fall outside of the range of the condence intervals generated by bootstrapping. This seems to most commonly be a problem when a population/locus contains a very little diversity and has a very small value standard deviation.

The distance matrices are pairwise comparisons of populations on a locus by locus basis. Matrices are provided for GST est , GST est , and Dest . Distance matrices are also calculated across loci by taking the harmonic mean.

3.3

How to Cite

Crawford NG. 2009. SMOGD: Software for the Measurement of Genetic Diversity. Molecular Ecology Resources. Accepted.

3.4

Usage

The website is pretty self-explanatory. But, briey: delete the sample data (control-A, delete), paste in your le, select the number of bootstrap replicates (max = 1000), and click submit. When the analysis nishes the page will refresh with html output and links for downloading tables.

4
4.1

Known Bugs
As of 08/09
Internal Server Error... This may occur for a number of reasons. Most likely you have managed to generate a divide by zero error. A common situation where this occurs is if you have a population that has missing genotypes for an entire locus. If your le works without bootstrapping, while bootstrapping induces the error, check to see if any of the estimated parameters are zero. If they are, youve found your problem. Generally missing data and bootstrapping dont get along if you have a lot of missing data or very small population sizes the chance of randomly generating populations that consist of entirely missing data increases. nan appears in results tables. Scipy is robust to divide by zero type errors and reports them as nan. Its possible to get divide by zero errors if populations are genetically similar. Bootstrapping may result in populations with no diversity especially if your populations are not particularly genetically diverse to begin with. If you submit a popgen format le with a single locus you cant have any spaces in the locus name. The popgen format expects a header line. The server on which I host SMOGD has a max CPU processing time of 2 minutes. So if your data-set takes longer than 2 minutes to analyze, youll get an internal server error. However, Ive successfully ran data sets with 600 individuals, 10 loci, and 20 populations so this should not be a common problem.

Works Cited

Acz el J, Dar oczy Z. 1975. On measures of information and their characterizations. Mathematics in Science and Engineering, vol. 115, Academic Press, New York, San Francisco, London, 1975, xii + 234 pp. Hedrick, PW. 2005. A Standardized Genetic Dierentiation Measure. Evolution 59(8), 1633-1638. Jost L. 2008. GST and its relatives do not measure dierentiation. Molecular Ecology 17(18), 4015-4026. [link] Nei M. 1973. Analysis of gene diversity in subdivided populations. Proceedings of the National Academy of Sciences, USA., 70(12, Pt 1), 3321-3323.

Nei M, Chesser RK. 1983. Estimation of xation indices and gene diversities. Annals of Human Genetics. 47(3), 253-259. Tsallis C, Bigatti E. 2004. Nonextensive statistical mechanics: A brief introduction. Continuum Mechanics and Thermodynamics, 16(3), 223-235.