DNA Slider, version 1.11 
January 5, 1998

 DNA Slider is a Macintosh program for performing a significance test of 
heterogeneity in the ratio of polymorphic sites to fixed differences in 
DNA sequence data. Help screens make the program fairly self-explanatory. 
It's basically the same idea as the test described in: McDonald, J.H. 
1996. Detecting non-neutral heterogeneity across a region of DNA sequence 
in the ratio of polymorphism to divergence. Molecular Biology and 
Evolution 13:253-260, except that several new statistical tests have been 
added which are often more powerful than the runs test. These new tests 
are described in McDonald, J.H. 1998. Improved tests for heterogeneity 
across a region of DNA sequence in the ratio of polymorphism to 
divergence. Molecular Biology and Evolution, in press. 
DNA Slider should run on any Macintosh. It runs fastest on a Mac with a 
PowerPC chip, and on older Macs it takes advantage of an FPU if one is 
present. I have no plans to port the program to any other platform. If 
you have data you want to analyze and don't have access to a Macintosh, 
let me know and I'll analyze your data (in return for effusive thanks in 
the acknowledgements, of course). I wrote the program using the 
CodeWarrior Pascal compiler, and you can look at the source code 
(http://udel.edu/~mcdonald/slidersource.html) if you want.
 To use the program, you must first align multiple sequences from one 
species and one sequence from a second species.  Ignore any areas aligned 
with gaps, where part of the sequence is not present in all individuals. 
Also ignore any amino acid replacement (non-synonymous) variation. Next, 
classify those sites which differ from the consensus sequence as 
polymorphisms or fixed differences, as illustrated below. 
Where a single nucleotide site has three or four bases, they are treated 
as if they were at adjacent sites. For example, a site which has C, T and 
G in species A would be counted as two polymorphisms.  Where a single 
nucleotide site has both a polymorphism and a fixed difference, they are 
treated as adjacent sites, and to be conservative they are put in the 
order that yields less extreme values of the test statistics.
Here are some example sequences, six from species A and one from species 
B, of a region 21 nucleotides long. They will look best in a fixed-width 
font such as Courier or Monaco. Dashes indicate identity to the first 
sequence, dots indicate the insertion of a gap in the alignment. 

site sp. A  sp. B

 1   T-----  - 
 2   G-----  C     fixed difference #1
 3   A-----  -
 4   A--CC-  -     polymorphism #1
 5   G-----  -
 6   C-----  .
 7   A-T---  .     gap, so ignore this polymorphism
 8   T-----  .
 9   G-----  -
10   C-G---  -     polymorphism #2
11   C-----  -
12   C----T  G     fixed difference #2 and polymorphism #3
13   A-----  -
14   G-----  A     fixed difference #3
15   T-----  -
16   C--.--  -
17   A--.--  T     gap, so ignore this fixed difference
18   G--.--  -
19   C-----  -
20   CT---G  -     polymorphisms #4 and #5
21   T-----  -

The input data set for the program would have the name of the gene (or 
other identifying information) on the first line, the number of variable 
sites on the second line, then one line for each variable site, giving 
the site number followed by a "p" or "f" for polymorphism or fixed 
difference. The file based on the above data would look like this:

nothing dehydrogenase
8
2 f
4 p
10 p
12 f
12 p
14 f
20 p
20 p

Because site 12 has both a polymorphism and a fixed difference, there are 
two possible orders. The order shown maximizes the number of runs, and 
thus is conservative for the runs test. It may also be conservative for 
the other tests, but both orders should be tried and the one yielding the 
less extreme values of the test statistics should be used. 
When running the program, different values of the recombination parameter 
(R=4Nr) should be tried, and to be conservative, the value giving the 
greatest probability of obtaining the observed or fewer number of runs by 
chance should be used. Because this probability varies little over a 
fairly broad range of R, it is unnecessary to try a large number of 
closely spaced R values. Setting R=2, 4, 8, 16, 32... seems to work well. 
Note that with higher R values, the time required for the simulations 
increases dramatically. 
 
========================================================= John H. 
McDonald
Department of Biological Sciences 
University of Delaware 
Newark, Delaware 19716

mcdonald@udel.edu