Welcome, Guest. Please Login
YaBB - Yet another Bulletin Board
 
  HomeHelpSearchLogin  
 
subtreeselectioncomparison program (Read 4336 times)
Ulf Sorhannus
Guest


subtreeselectioncomparison program
Sep 11th, 2004 at 6:20am
 
Hello Everybody,

I am using the subtreeselectioncomparison program to test for subtree specific selective pressure. It is difficult to find, based on the list of nodes, the specific internal node at which the subtree of interest is rooted. My impression is that the node of interest is not on the list that is displayed in subtreeselectioncomparison program?? However, I may be wrong about that?? Is there a way (i.e. in HYPHY) to display all the node numbers of tree - in that way one could easily pick the node of interest. There are 43 sequences in the data matrix.

Thank you very much in advance!

Ulf  
Back to top
« Last Edit: Sep 13th, 2004 at 10:08am by Sergei »  
 
IP Logged
 
Sergei
YaBB Administrator
*****
Offline


Datamonkeys are forever...

Posts: 1658
UCSD
Gender: male
Re: subtreeselectioncomparison program
Reply #1 - Sep 13th, 2004 at 10:32am
 
Dear Ulf,

There are two ways to inspect trees or specify subtrees a priori:

(1). Edit the tree string to give an internal node you need to label a custom name. For example, if you want to refer to the subtree ((1,2),3) in the following tree:  (((1,2),3),(4,5),6), you can explicitly edit the tree string as such: (((1,2),3)MySubtree,(4,5),6). Naturally, you can call 'MySubtree' anything you want, but in the subtreeselectioncomparison analysis dialog, you will be able to see 'MySubtree' as a choice.

(2). I have edited the subtreeselectioncomparison.bf analysis so that it now displays a graphic of the tree with all the nodes labeled behind the selection dialog. You can move the selection dialog to the side and inspect the tree for the name of the subtree-rooting internal node you wish to use. This only works on Mac and Windows. The modified analysis will be included in today's and later HyPhy builds. You can also download it directly Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login Login and replace the corresponding file in 'TemplateBatchFiles'.

Hope this helps and thanks for your feedback!

Sergei
Back to top
 

Associate Professor
Division of Infectious Diseases
Division of Biomedical Informatics
School of Medicine
University of California San Diego
WWW WWW  
IP Logged
 
Ulf
Guest


Re: subtreeselectioncomparison program
Reply #2 - Sep 13th, 2004 at 12:03pm
 
Dear Sergei,

Thank you very much for your help! I tried the first approach and it works. Ulf
Back to top
 
 
IP Logged
 
ulf
Guest


Re: subtreeselectioncomparison program
Reply #3 - Sep 13th, 2004 at 4:40pm
 
Dear Sergei,

Below is an example (portion) of the output  from the subtreeselection comparison analysis. I am not sure how to interpret the statistically significant results?  Does it mean that the subclade that was chosen for the analysis show positive selection and the rest of the tree does not (or vice versa)? There are several of the significant sites (*) in the output.  Thanks a million! Ulf

Example output:

Site   42 dN Tree= 14.220 dN Clade  0.000 dS =  1.468 Joint dN  1.518 Joint dS  1.477 LRT  9.318 p-value =  0.0023 *
Site   43 dN Tree=  6.559 dN Clade  0.000 dS =  2.903 Joint dN  0.655 Joint dS  2.897 LRT  4.676 p-value =  0.0306 *
Site   44 dN Tree=  0.000 dN Clade  1.792 dS =  0.000 Joint dN  1.541 Joint dS  0.000 LRT  0.523 p-value =  0.4694
Site   45 dN Tree=  0.000 dN Clade  0.000 dS =  2.050 Joint dN  0.000 Joint dS  2.050 LRT -0.000 p-value =  1.0000
Site   52 dN Tree= 18.713 dN Clade  0.000 dS =  1.878 Joint dN  2.104 Joint dS  1.918 LRT  9.117 p-value =  0.0025 *
Back to top
 
 
IP Logged
 
Sergei
YaBB Administrator
*****
Offline


Datamonkeys are forever...

Posts: 1658
UCSD
Gender: male
Re: subtreeselectioncomparison program
Reply #4 - Sep 14th, 2004 at 8:12am
 
Dear Ulf,

All pairwise site-by-site comparative analyses are very similar in these basic concepts:

(1). First, we partition the tree into two groups of branches: "Clade" and the rest of the tree ("Tree"). "Clade" could be a subtree (your case), or any other collection of branches, e.g. all internal branches.

(2). Having fitted all the nuisance parameters (branch lengths, nucleotide biases etc) using the entire alignment, we hold those fixed for site-by-site evatuations.

(3). At every site s, we fit a synonymous rate alpha_s (dS) and two separate non-synonymous rates: beta_s^1 (dN clade) and beta_s^2 (dN tree). subtreeselectioncomparison.bf file then tests whether of not beta_s^1=beta_s^2, using a 1-df LRT. Thus, a significant result tells you that selective pressures between the "clade" and the rest of the tree are different. However this could be due to many reasons, e.g. one is negative and the other is positive, or one is positive and the other is very positive, etc. You can use estimates of dS and dN to interpret the direction of changes, but it is not a rigorous comparison. For example:

"Site   42 dN Tree= 14.220 dN Clade  0.000 dS =  1.468 Joint dN  1.518 Joint dS  1.477 LRT  9.318 p-value =  0.0023 *"

The dN/dS in the clade is 0 (0/1.468 ), while in the rest of the tree it is about 10 (14.22/1.468 ). When dN/dS is forced to be the same accross the entire tree, the value is about 1 (1.518/1.477). This suggests that while on average the site is evolving neutrally, there is strong negative selection in the clade and strong positive selection in the rest of the tree.

You can use the analysis 'subtreespecificselection' to test whether or not beta_s^1=alpha_s, and thus check for non-neutral evolution in a clade rigorously.

HTH,
Sergei
Back to top
 

Associate Professor
Division of Infectious Diseases
Division of Biomedical Informatics
School of Medicine
University of California San Diego
WWW WWW  
IP Logged
 
Ulf Sorhannus
Guest


You can use Re: subtreeselectioncomparison program
Reply #5 - Sep 14th, 2004 at 1:27pm
 
Dear Sergei,

"You can use the analysis 'subtreespecificselection' to test whether or not beta_s^1=alpha_s, and thus check for non-neutral evolution in a clade rigorously. "

Could you fill me in on how to do the above test using the subtreeselectioncomparison program....?Thanks! Ulf
Back to top
 
 
IP Logged
 
Sergei
YaBB Administrator
*****
Offline


Datamonkeys are forever...

Posts: 1658
UCSD
Gender: male
Re: subtreeselectioncomparison program
Reply #6 - Sep 15th, 2004 at 2:08pm
 
Dear Ulf,

"Could you fill me in on how to do the above test using the subtreeselectioncomparison program....?Thanks! Ulf"

I realized that you could not actually do this test in any of the standard analysis files. I have added the option to the QuickSelectionDetection.bf analysis. Download the updated file from Multimedia File Viewing and Clickable Links are available for Registered Members only!!  You need to Login Login and place it in "TemplateBatchFiles". Then go through the analysis options, choosing "Two-rate ALRS" when prompted for the method. You will see a dialog for which branches to test the selection on, and you can choose the subtree you want. What will now be tested is whether dN/dS = 1 is along that subtree (dN/dS is free to vary along the rest of the tree).

HTH,
Sergei
Back to top
 

Associate Professor
Division of Infectious Diseases
Division of Biomedical Informatics
School of Medicine
University of California San Diego
WWW WWW  
IP Logged