HyPhy message board - How to confirm GARD result?

kevintz YaBB Newbies Offline Feed your monkey! Posts: 15	Re: How to confirm GARD result? Reply #30 - Jan 13^th, 2009 at 2:32pm Dear Sergi, I get it. The program runs well. Really thanks. One more basic question, does the AICc value refer to that from GARD result profile, or from the KH test result file? I find both of them include a AICc, but with different value. Which one is better? Best regards, kevintz
Back to top	IP Logged

Sergei

YaBB Administrator

Offline

Datamonkeys are forever...

Posts: 1658
UCSD
Gender: male

Re: How to confirm GARD result?
Reply #31 - Jan 14^th, 2009 at 12:35pm

Dear kevintz,

KH test file should always have a somewhat better AICc than the GARD file, because GARD only estimates several model parameters (e.g. nucleotide biases and site-to-site rate variation parameters) only once (assuming no recombination) to speed up the run and the KH analysis estimates those parameters assuming recombination. The difference in AICc should not be very large however.

Sergei

Back to top

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

WWW

IP Logged

kevintz YaBB Newbies Offline Feed your monkey! Posts: 15	Re: How to confirm GARD result? Reply #32 - Jan 14^th, 2009 at 11:01pm THanks
Back to top	IP Logged

kevintz

YaBB Newbies

Offline

Feed your monkey!

Posts: 15

Re: How to confirm GARD result?
Reply #33 - Jan 18^th, 2009 at 10:41am

Sergei wrote on Jan 13^th, 2009 at 8:07am:

kevintz wrote on Jan 13^th, 2009 at 1:42am:

Thanks for your reply.
the following is the file.

Another p-value (partition 5, 0.0014) becomes greater than the threshold of 0.01 following a Bonferroni correction for multiple tests; hence that partition is judged insignificantly supported.

Cheers,
Sergei

Dear Sergei,
Sorry to bother you again.

I read your article on journal Gene 397 (2007) 38-50, entitled "Evolution of the interferon alpha gene family in eutherian mammals", and found Table 1 refered to KH tesing with all the P-value <0.01. While in my data set, the KH testing resluts show the p-value much smaller as far as 0.001. As you said, Bonferroni correction might be performed. So, my question is does the KH test doing Bonferroni correction automatically? if not, how? if yes, why not to refer in the article?

Thanks a million.

Best regards,
Tu Zeng

Back to top

IP Logged

Sergei YaBB Administrator Offline Datamonkeys are forever... Posts: 1658 UCSD Gender:	Re: How to confirm GARD result? Reply #34 - Jan 18^th, 2009 at 5:33pm Dear Tu Zeng, The numbers in the paper are quoted following a Bonferroni correction. The KH test provides raw numbers for the p-values, but includes a Bonferroni correction when counting the number of significant partitions. HTH, Sergei
Back to top	Associate Professor Division of Infectious Diseases Division of Biomedical Informatics School of Medicine University of California San Diego WWW IP Logged

kevintz

YaBB Newbies

Offline

Feed your monkey!

Posts: 15

Re: How to confirm GARD result?
Reply #35 - Jan 20^th, 2009 at 2:38am

Sergei wrote on Jan 18^th, 2009 at 5:33pm:

Dear Tu Zeng,

The numbers in the paper are quoted following a Bonferroni correction. The KH test provides raw numbers for the p-values, but includes a Bonferroni correction when counting the number of significant partitions.

HTH,
Sergei

Thanks it helps me a lot.
And How I can know the value of the Bonferroni correction through Hyphy or KH test? whether it set to 0.05 or 0.01 after KH test?
Thanks a lot.
Best,
kevintz.

PS: tuzeng is my name, sorry for confusion.

Back to top

IP Logged

Sergei YaBB Administrator Offline Datamonkeys are forever... Posts: 1658 UCSD Gender:	Re: How to confirm GARD result? Reply #36 - Jan 20^th, 2009 at 9:27am Dear tuzeng, The processor file considers a Bonferroni-corrected p<=0.01 to be significant. Cheers, Sergei
Back to top	Associate Professor Division of Infectious Diseases Division of Biomedical Informatics School of Medicine University of California San Diego WWW IP Logged

Sundy

YaBB Newbies

Offline

Feed your monkey!

Posts: 33

Re: How to confirm GARD result?
Reply #37 - Feb 12^th, 2009 at 7:48am

kevintz wrote on Jan 18^th, 2009 at 10:41am:

Dear Sergei,
Sorry to bother you again.

I read your article on journal Gene 397 (2007) 38-50, entitled "Evolution of the interferon alpha gene family in eutherian mammals", and found Table 1 refered to KH tesing with all the P-value <0.01. While in my data set, the KH testing resluts show the p-value much smaller as far as 0.001. As you said, Bonferroni correction might be performed. So, my question is does the KH test doing Bonferroni correction automatically? if not, how? if yes, why not to refer in the article?

Thanks a million.

Best regards,
Tu Zeng

Dear Sergei,

I am very interested in the table 1 of above paper, and I also several basice questions aobut it.

1. Where can I find the AICc improvement in my KH test result file? for example from my attached file. Is it Delta AIC = 280.92?
2. Where can I find the LHS vs. RHS and RHS vs LHS. For example from my attached file.

Thank you very much!!

Sundy

Due to the character limitation, I deleted the "Tree aTree=[...]" and attached the entire file.

****************
Loaded 5 partitions

Sequences :10
Sites :3666
Variable :539

f(A) = 0.285291
f(C) = 0.227325
f(G) = 0.205476
f(T) = 0.281908

c-AIC = 17279.5
Log Likelihood = -8614.56089283971;
Shared Parameters:
GT=0.221445=0.221445
CT=1.90112=1.90112
CG=0.10546=0.10546
AT=0.471935=0.471935
AC=0.281374=0.281374
alpha=0.0102205=0.0102205
betaQ=3.72107=3.72107
betaP=4.31789=4.31789

Tree givenTree=(((Bat_SARS_273_DQ648856:0.0496989,((Bat_SARS_279_DQ648857:0.0282901,B
at_SARS_Rp_DQ071615:0.0244375)Node5:0.00575514,Bat_SARS_HKU3_DQ022305:0.0359314)
Node4:0.0267917)Node2:0.0374558,3_HP03E_GZ02:0.000232858)Node1:0.000864494,(130_
PC03_SZ13:0.000540745,(110_PC04_PC4_136:0.00192447,106_HP04_GZ0402:0.00137285)No
de12:0.0016568)Node10:0.00110442,(15_HP03M_BJ02:0.000823405,31_HP03L_Tor2:0.0002
70703)Node15:0.000270703);
Mean divergence : 6.49111%

Fitting a single-tree, multiple partition model
Log Likelihood = -8578.32862962121;
Shared Parameters:
S_1=0.92014
S_2=0.382852
S_3=0.560115
S_4=0.375399
betaP=4.31789=4.31789
betaQ=3.72107=3.72107
alpha=0.0102205=0.0102205
AC=0.281374=0.281374
AT=0.471935=0.471935
CG=0.10546=0.10546
CT=1.90112=1.90112
GT=0.221445=0.221445

Tree tree_0=

Fitting a mutilple tree, multiple partition model
Log Likelihood = -8403.83189116324;
Shared Parameters:
betaP=4.31789=4.31789
betaQ=3.72107=3.72107
alpha=0.0102205=0.0102205
AC=0.281374=0.281374
AT=0.471935=0.471935
CG=0.10546=0.10546
CT=1.90112=1.90112
GT=0.221445=0.221445

Tree tree_0=

Versus the single partition model: c-AIC = 16998.6
Delta AIC = 280.92

Versus the single tree/multiple partition model: Delta AIC = 216.577

Partition 1 : 381 sites
Partition 2 : 343 sites
Partition 3 : 1080 sites
Partition 4 : 722 sites
Partition 5 : 1140 sites

Fitting tree 1 to partition 1
Log Likelihood = -1181.64638188419;
Shared Parameters:
betaP=4.31789=4.31789
betaQ=3.72107=3.72107
alpha=0.0102205=0.0102205
AC=0.281374=0.281374
AT=0.471935=0.471935
CG=0.10546=0.10546
CT=1.90112=1.90112
GT=0.221445=0.221445

Tree aTree=

Fitting tree 2 to partition 1
Log Likelihood = -1195.56564111684;
Shared Parameters:
betaP=4.31789=4.31789
betaQ=3.72107=3.72107
alpha=0.0102205=0.0102205
AC=0.281374=0.281374
AT=0.471935=0.471935
CG=0.10546=0.10546
CT=1.90112=1.90112
GT=0.221445=0.221445

Tree aTree=

KH Testing partition 1
Tree 2 base LRT = 27.8385. p-value = 0.025

Fitting tree 1 to partition 2
Log Likelihood = -977.448251979626;
Shared Parameters:
betaP=4.31789=4.31789
betaQ=3.72107=3.72107
alpha=0.0102205=0.0102205
AC=0.281374=0.281374
AT=0.471935=0.471935
CG=0.10546=0.10546
CT=1.90112=1.90112
GT=0.221445=0.221445

Tree aTree=

Fitting tree 2 to partition 2
Log Likelihood = -968.982471135205;
Shared Parameters:
betaP=4.31789=4.31789
betaQ=3.72107=3.72107
alpha=0.0102205=0.0102205
AC=0.281374=0.281374
AT=0.471935=0.471935
CG=0.10546=0.10546
CT=1.90112=1.90112
GT=0.221445=0.221445

Tree aTree=

Fitting tree 3 to partition 2
Log Likelihood = -972.312843728842;
Shared Parameters:
betaP=4.31789=4.31789
betaQ=3.72107=3.72107
alpha=0.0102205=0.0102205
AC=0.281374=0.281374
AT=0.471935=0.471935
CG=0.10546=0.10546
CT=1.90112=1.90112
GT=0.221445=0.221445

Tree aTree=

KH Testing partition 2
Tree 1 base LRT = 16.9316. p-value = 0.0001
Tree 3 base LRT = 6.66075. p-value = 0.0001

Fitting tree 2 to partition 3
Log Likelihood = -2258.19915763715;
Shared Parameters:
betaP=4.31789=4.31789
betaQ=3.72107=3.72107
alpha=0.0102205=0.0102205
AC=0.281374=0.281374
AT=0.471935=0.471935
CG=0.10546=0.10546
CT=1.90112=1.90112
GT=0.221445=0.221445

Tree aTree=

Fitting tree 3 to partition 3
Log Likelihood = -2244.90399830231;
Shared Parameters:
betaP=4.31789=4.31789
betaQ=3.72107=3.72107
alpha=0.0102205=0.0102205
AC=0.281374=0.281374
AT=0.471935=0.471935
CG=0.10546=0.10546
CT=1.90112=1.90112
GT=0.221445=0.221445

Tree aTree=

Fitting tree 4 to partition 3
Log Likelihood = -2285.99836857422;
Shared Parameters:
betaP=4.31789=4.31789
betaQ=3.72107=3.72107
alpha=0.0102205=0.0102205
AC=0.281374=0.281374
AT=0.471935=0.471935
CG=0.10546=0.10546
CT=1.90112=1.90112
GT=0.221445=0.221445

Tree aTree=

KH Testing partition 3
Tree 2 base LRT = 26.5903. p-value = 0.0001
Tree 4 base LRT = 82.1887. p-value = 0.0001

Fitting tree 3 to partition 4
Log Likelihood = -1640.47340028827;
Shared Parameters:
betaP=4.31789=4.31789
betaQ=3.72107=3.72107
alpha=0.0102205=0.0102205
AC=0.281374=0.281374
AT=0.471935=0.471935
CG=0.10546=0.10546
CT=1.90112=1.90112
GT=0.221445=0.221445

Tree aTree=

Fitting tree 4 to partition 4
Log Likelihood = -1632.78789828693;
Shared Parameters:
betaP=4.31789=4.31789
betaQ=3.72107=3.72107
alpha=0.0102205=0.0102205
AC=0.281374=0.281374
AT=0.471935=0.471935
CG=0.10546=0.10546
CT=1.90112=1.90112
GT=0.221445=0.221445

Tree aTree=

Fitting tree 5 to partition 4
Log Likelihood = -1639.94300504171;
Shared Parameters:
betaP=4.31789=4.31789
betaQ=3.72107=3.72107
alpha=0.0102205=0.0102205
AC=0.281374=0.281374
AT=0.471935=0.471935
CG=0.10546=0.10546
CT=1.90112=1.90112
GT=0.221445=0.221445

Tree aTree=

KH Testing partition 4
Tree 3 base LRT = 15.371. p-value = 0.0001
Tree 5 base LRT = 14.3102. p-value = 0.0001

Fitting tree 4 to partition 5
Log Likelihood = -2375.98828232171;
Shared Parameters:
betaP=4.31789=4.31789
betaQ=3.72107=3.72107
alpha=0.0102205=0.0102205
AC=0.281374=0.281374
AT=0.471935=0.471935
CG=0.10546=0.10546
CT=1.90112=1.90112
GT=0.221445=0.221445

Tree aTree=

Fitting tree 5 to partition 5
Log Likelihood = -2375.51119382618;
Shared Parameters:
betaP=4.31789=4.31789
betaQ=3.72107=3.72107
alpha=0.0102205=0.0102205
AC=0.281374=0.281374
AT=0.471935=0.471935
CG=0.10546=0.10546
CT=1.90112=1.90112
GT=0.221445=0.221445

Tree aTree=

KH Testing partition 5
Tree 4 base LRT = 0.954177. p-value = 0.0001
A total of 3/4 significant couplings
Mean splits identify: 0.26

***********************

Back to top

Multimedia File Viewing and Clickable Links are available for Registered Members only!! You need to

Login (2 KB | )

IP Logged

Sergei

YaBB Administrator

Offline

Datamonkeys are forever...

Posts: 1658
UCSD
Gender: male

Re: How to confirm GARD result?
Reply #38 - Feb 12^th, 2009 at 2:29pm

Dear Sundy,

The first Delta AIC is the one you want (the second tests a slightly different model, and it is not mentioned in the paper).

The LHS vs RHS values are reported in fragments like this:

Code:

KH Testing partition 1
Tree 2 base LRT = 27.8385. p-value = 0.025

This specific output states that the KH p-value for the LHS (partition 1) vs the tree from the RHS (partition 2) is 0.025. Hence this would be the LHS v RHS value from the paper. Looking further down, you will find

Code:

KH Testing partition 2
Tree 1 base LRT = 16.9316. p-value = 0.0001

which states that the RHS (partition 2) has KH p-value of 0.0001 vs the tree from the partition 1 (LHS) -- this is the RHS vs LHS value.

If you have multiple breakpoints, a pair of p-values like this will be reported for each breakpoint.

HTH,
Sergei

Back to top

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

WWW

IP Logged

Sundy YaBB Newbies Offline Feed your monkey! Posts: 33	Re: How to confirm GARD result? Reply #39 - Feb 12^th, 2009 at 2:59pm Dear Sergei, Thank you so much! I got it Cheers, Sundy
Back to top	IP Logged

	Welcome, Guest. Please Login