HyPhy message board | |
http://www.hyphy.org/cgi-bin/hyphy_forums/YaBB.pl
HYPHY Package >> HyPhy feedback >> hyphybook2007.pdf http://www.hyphy.org/cgi-bin/hyphy_forums/YaBB.pl?num=1290902890 Message started by hs on Nov 27th, 2010 at 4:08pm |
Title: Re: hyphybook2007.pdf Post by hs on Dec 2nd, 2010 at 10:17am
How to 'export the best fitting nucleotide model by writing the likelihood function object to a NEXUS-formatted file' from a command-line?
When I selected '(2):[Yes] Save each of the 203 files to a separate file' for '|Save each of the 203 fits?|' in NucModelCompare.bf, and then selected '(2):[Restore] Restore an earlier nucleotide fit' for '|New/Restore|' and gave the file in QuickSelectionDetection.bf, HYPHY printed the following error messages. ########## Error: DataSet/DataSetFilter nucData has not been initialized Function call stack 1 : Harvest Frequencies into matrix positionFrequencies from DataSet nucData with unit size = 3 with atom size = 1 with position specific flag = 1 Partition: ########## [references] hyphybook2007.pdf # 2.3.7 Estimating nucleotide biases It is usually a good idea to export the best fitting nucleotide model by writing the likelihood function object to a NEXUS-formatted file, which is carried out in HyPhy by selecting the menu option ‘Analysis > Results > Save Results > Export Analysis’. In section 2.6, we will describe how to import the fitted nucleotide model from this file into a selection analysis. # 2.6.1 Preliminary analysis set-up (ii) New/Restore — Restore nucleotide model parameter estimates from a previous analysis. Choosing ‘Restore’ requires you to select a file that contains a previously exported likelihood function, which will also contain the sequence data and tree. This option allows you to iteratively run the batch file under different settings without having to re-optimize the same nucleotide model for every session. It also provides a convenient means of importing the best-fitting nucleotide model from the automated model selection procedure implemented in the batch file NucModelCompare.bf (see section 2.3.7). |
Title: Re: hyphybook2007.pdf Post by hs on Dec 3rd, 2010 at 1:15pm
Are there a list of 6 character for all standard or ‘named’ models of nucleotide substitution (e.g:010010 defines HKY85)?
Does 012345 define General Reversible Model? |
Title: Re: hyphybook2007.pdf Post by Sergei on Dec 3rd, 2010 at 3:06pm
Hi hs,
F81 : 000000 HKY85 : 010010 TrN93 : 010020 REV : 012345 Cheers, Sergei |
Title: Re: hyphybook2007.pdf Post by Sergei on Dec 3rd, 2010 at 3:07pm
Hi hs,
F81 : 000000 HKY85 : 010010 TrN93 : 010020 REV : 012345 Cheers, Sergei |
Title: Re: hyphybook2007.pdf Post by hs on Dec 8th, 2010 at 11:17am
In SLAC and FEL, selecting '(3):[Estimate] Estimate from data with branch corrections(slower)' rather than '(5):[Estimate dN/dS only]' at 'dN/dS bias parameter options' tend to produce dN/dS = 0/0 rather than dN/dS = inf (∞)?
The sites with dN/dS = inf (∞) can be declared to be under positive selection if p-value < 0.05? |
Title: Re: hyphybook2007.pdf Post by Sergei on Dec 10th, 2010 at 3:19pm
Hi hs,
Can you paste in the console output for your file when you see dN/dS =0/0? I am not sure I understand what the problem is. Sergei |
Title: Re: hyphybook2007.pdf Post by hs on Dec 10th, 2010 at 4:22pm
I wonder if I could conclude that the site with p-value < 0.05 are under positive selection even if the site has dS = 0?
dN/dS =0/0 is not the console output but the statement in hyphybook2007.pdf ('β /α may assume the undefined value 0/0 or ∞' as pasted below). ########## 2.6.4 Fixed effects likelihood (FEL) dN/dS - the ratio of β /α for the branches of interest (could be undefined or infinite if α is estimated to be 0) 2.5.1 Fitting a local codon model in HyPhy It is not unusual for estimates of α or β to converge to zero for one or more branches in a tree. Clearly, this outcome implies that there is insufficient sequence variation to support a branch of non-zero length in that part of the tree. If this occurs, then local estimates of the ratio β /α may assume the undefined value 0/0 or ∞. As a result, the tree will not be displayed properly in the HyPhy tree viewer window when it is scaled to β /α. Selection analyses performed on such trees, however, will not be affected by these poorly-defined branch lengths. ########## |
Title: Re: hyphybook2007.pdf Post by Sergei on Dec 10th, 2010 at 4:56pm
Hi hs,
hs wrote on Dec 2nd, 2010 at 10:17am:
It would appear that the likelihood function output generated by NucModelCompare.bf is not directly importable by QuickSelectionDetection.bf. The text in the book is in error. Thanks for pointing this out. If you replace NucModelCompare.bf with the file that I have attached, this should be fixed. Thanks for the bug report! Sergei http://www.hyphy.org/cgi-bin/hyphy_forums/YaBB.pl?action=downloadfile;file=NucModelCompare.bf (24 KB | )
|
HyPhy message board » Powered by YaBB 2.5.2! YaBB Forum Software © 2000-2024. All Rights Reserved. |