codeml stops running after first model without any error message

[Maude Jacquot]

 Hi all,

I want to use PAML to carry out the selection analysis of 43 sequences. 

I set up an analysis to run the M0, M1, M2 and M3 models in codeml. The analysis starts to run well for the first model (M0) and then stops to run without any error message.  

I tried both the command line and PAML-X tools. Results obtained are the same. 

It looks like the inf file is not normal and I have this extra file called 4fold.nuc which is not created by the example provided by developers.

I attached here my config file and the outputs. I could send the inputs privately too.

Does anyone have a idea? Any thought would be much appreciated.

All the best,

Maude

[Ziheng]

the tree file is missing so I can't tell what may be wrong. 

the output file ends with

"Model 0: one-ratio"

without results, which may indicate a problem with the tree file.

are you sure there were no error messages on the monitor?

ziheng

[Maude Jacquot]

Hi Ziheng,

Thank you for your answer. Here are my inputs files as well a a screenshot of the monitor. Hope this helps, I definitely can't figure out what's happening.

All the best,

Maude

[Sungmin Lee]

Hi, 

Species 0.78? in the bottom of your terminal is actually an error message. It looks like you used some sort of program that generates nodal/branch support for the tree estimation. Since the node is defined and there is no associated sequence, paml will not proceed the analysis unless you remove those. You can write a python script to deal with them.

Best,

Sung

[Maude Jacquot]

Hi Sung,

Thank you very much, this solved my problem.

All the best,

Maude

[Maude Jacquot]

Hi again,

Thanks again for your advise: I've been able to run PAML to carry out the selection analysis of sequences. I set up an analysis to run the M0, M1, M2 and M3 models in codeml. I'm now struggling with the interpretation of the outputs. I want to determine which model best fit my data. To do this, I assume I have to compare the log likelihood values found in the mlc file:

Model 0: -9497.825012

Model 1: -9480.796282

Model 2: -9480.796285

Model 3: -9480.474092

Here, the model 3 (discrete) seems to be the best, right?

What do you think about the really small differences between M1, 2 and 3 lnL? Can I be confident in saying that one of these models is better than another one?

Thanks for your help,

Maude

[Sungmin Lee]

Hi Maude, 

You can compare M3 with M0 and M2 with M1. Do a LRT to see whether the null model is rejected or not. I think M3 shows a variation rather than the positive selection, this was mentioned in one of Dr. Yang paper as far as I remember.

Cheers,

Sung