Hello everyone,

I'm trying to compare the results of an analysis of positive selection on HyPhy versus datamonkey website. The problem I have is that both give me completely different results.

With the following options in Hyphy, I get the following results:

First get the tree with:

(9) Phylogeny Reconstruction
(3) Perform a phylogeny reconstuction for nucleotide, protein or codon data with user-selectable models using the method of neighbor joining.
(1):[Distance formulae] Use one of the predefined distance measures based on data comparisons. Fast.
(2):[Codon] Codon (several available genetic codes).
(1):[Universal] Universal code. (Genebank transl_table=1).
Please choose a codon data file::/home/.....
(2):[Force Zero] Negative Branch Lengths are Forced to 0.
(1):[Nei_Gojobori] Nei and Gojobori (1986) method.
(1):[Joint p-distance(Default)] Observed N/E[N]+Observed S/E[S]

And then, my analysis of positive selection:

(1):[Universal] Universal code. (Genebank transl_table=1).
(1):[New Analysis] Perform a new analysis.
Please specify a codon data file:: /home/.....
(1):[Default] Use HKY85 and MG94xHKY85.
Please select a tree file for the data::/home/.....
/Save nucleotide model fit to::/home/....
(1):[Neutral] dN/dS=1
(1):[Single Ancestor Counting] Use the most likely ancestor state
Branch Corrections Factor (<0 to estimate):-1
(1):[Full tree] Analyze the entire tree
(1):[Averaged] All possible resolutions are considered and averaged.
(1):[Approximate] Use the approximate extended binomial distribution (fast)
Significance level for a site to be classified as positively/negatively selected?0.01

******* FOUND 6 POSITIVELY SELECTED SITES ********

+--------------+--------------+--------------+--------------+
| Index        | Site Index   | dN-dS        | p-value      |
+--------------+--------------+--------------+--------------+
|            1 |   266.000000 |    16.975860 |     0.004127 |      
+--------------+--------------+--------------+--------------+
|            2 |   711.000000 |    17.775800 |     0.001779 |      
+--------------+--------------+--------------+--------------+
|            3 |   871.000000 |    15.384795 |     0.007866 |      
+--------------+--------------+--------------+--------------+
|            4 |   933.000000 |    13.977618 |     0.008030 |      
+--------------+--------------+--------------+--------------+
|            5 |   963.000000 |    15.981690 |     0.008440 |      
+--------------+--------------+--------------+--------------+
|            6 |   964.000000 |    20.738813 |     0.001250 |      
+--------------+--------------+--------------+--------------+


(2):[Export to File] Output is spooled to a tab separated file.
Export tab separated data to::/home/....
(1):[Skip] Skip the estimation of number of dS and dN rate classes

Now, with the following options in datamonkey, I get the following results:

Read 35 sequences and 1367 codon alignment columns and 1 partitions.
Nucleotide composition
    A 22.0984%
    C 29.0403%
    G 26.9878%
    T 21.8734%
Method: SLAC
Use this tree set:Neighbor Joining Tree
Define a custom (or choose a "named" HKY85) nucleotide substitution bias model
Global dN/dS value is: Neutral      1.0
Handling ambiguities: Averaged
Significance Level (p-value/Bayes Factor/posterior probability): 0.01

Data summary
35 sequences with 1 partition
Partition 1: 1367 codons 17.0656 subs/site
Nucleotide Model(010010) Fit Results
Log(L) = -102728.673
Relative substitution rates

     A      C            G            T
A      *      0.890379      1            0.890379
C      -      *            0.890379      1
G      -      -            *            0.890379
T      -      -            -            *

Codon Model Fit Results (processor time taken: 100.97 seconds)
Log(L) = -101613  mean dN/dS = 1

Found 7 positively selected sites ( significance level 0.01)

Codon      dN-dS      Normalized dN-dS      p-value      
77        21.7536      0.424902      0.00730707      
175        33.3654      0.651709      0.00675704      
434        25.8345      0.504611      0.00679637      
738        25.4569      0.497236      0.00283785                  
1095   31.5538      0.616323      0.0021509      
1143   24.7542      0.48351      0.00337522      
1198   26.4843      0.517304      0.00461045      

Well, as you can see, in both analysis I get different results. I mean that in each tests I get different sites under positive selection. I have done the same exercise with the sample file that appears on the website of datamonkey (Influenza A H5N1 hemagluttinin) and in that case both analysis gave me the same results (this was what I expected for my records). For this reason I believe that perhaps there is something wrong with my alignment. I attached my file to see if someone can help me. Another thing that comes to mind is that the options that I'm using to generate the tree in Hyphy maybe are differents to those used in datamonkey website.

Well, I hope someone can help me.

Gabriel

Dear Sergei,

Thank you very much for your reply. I used the nucleotide NJ (TN93) trees to perform my hyphy analysis and also eliminated the poorly aligned positions and divergent regions of my alignments with Gblocks and the results were much more agreement.

I take this ocation to ask you, what parameter do you recommend for testing positive selection with sequences from widely divergent species?

Gabriel