Models supported in epa-ng

[kanellopal]

Hey Pierre.

Thanks a lot for developing this software. One question I have is which AA substitution models are supported in epa-ng?

I used RAxML-NG v2.0.1 to optimise some parameters for a tree built with iqtree3 with the Q.PFAM+R10 model. I get the .BestModel file which looks like this:
Q.PFAM+R10{0.051755/0.163096/0.288424/0.431771/0.599218/0.800573/1.053245/1.393043/1.916291/3.302585}{0.129458/0.108723/0.121044/0.116191/0.104157/0.099554/0.107772/0.071831/0.070528/0.070742}, noname = 1-603

However when I try to use it with the --model argument in epa-ng v0.3.8 with
epa-ng --ref-msa final_ureC_nogap095.afa --tree ureC_raxML.raxml.bestModel_rooted.tree --query ureC_fragments_10000.fna --model ureC_raxML.raxml.bestModel -T 12 or
epa-ng --ref-msa final_ureC_nogap095.afa --tree ureC_raxML.raxml.bestModel_rooted.tree --query ureC_fragments_10000.fna --model Q.pfam+R10{0.051755/0.163096/0.288424/0.431771/0.599218/0.800573/1.053245/1.393043/1.916291/3.302585}{0.129458/0.108723/0.121044/0.116191/0.104157/0.099554/0.107772/0.071831/0.070528/0.070742} -T 12

I receive the following error

INFO Selected: Output dir: ./
INFO Selected: Query file: ureC_fragments_10000.fna
INFO Selected: Tree file: ureC_raxML.raxml.bestModel_rooted.tree
INFO Selected: Reference MSA: final_ureC_nogap095.afa
INFO Selected: Automatic switching of use of per rate scalers
INFO Selected: Preserving the root of the input tree
INFO Selected: Specified model file: ureC_raxML.raxml.bestModel
libc++abi: terminating due to uncaught exception of type std::runtime_error: Invalid model name: Q.PFAM
zsh: abort epa-ng --ref-msa final_ureC_nogap095.afa --tree --query --model -T 12

I suspect the Q.pfam model may not be supported in epa-ng even though it is included in RAxML-NG v2.0.1 and iqtree. But maybe I am doing something else wrong, so I would like your opinion.

Cheers,

Alex

[lczech]

Hi Alex @kanellopal,

the issue you are seeing is because EPA-ng does not (yet) include the most recent models from RAxML-ng. Hopefully we will make them available here soon. In the meantime, I've already added them to my fork of EPA-ng (here), but just ad-hoc, so I did not further test if the changes have any other effects. Could you maybe try with that fork, compile it from scratch, and see if that solves your problem?

Thanks, and greetings from Denmark to Sweden
Lucas

[kanellopal]

Hey Lucas. Thanks for the swift response. I will try and let you know, I have to play around with some cmake versions since I have one > 3.5 and it does not want to build epa-ng

[lczech]

Hey Alex @kanellopal,

over the weekend @pierrebarbera made a new release as well here now, which contains the new changes. It does not seem to have been caught by conda yet, probably because it's a pre-release, but maybe this one works better for you.

Let us know how things are going, and so long
Lucas

[kanellopal]

Hey @lczech

Thank you very much for the heads-up. Sorry for the radio silence, I had trouble compiling your fork, but I finally managed to do it this morning and I am testing right now actually.

I'll let you know if something seems broken.

Cheers,

Alex

[kanellopal]

From a first reading it seems to be working. I managed to get placements for some query sequences. The only question I have is if it is possible to specify the rate heterogeneity parameters. Right now I am using:

epa-ng --ref-msa final_ureC_nogap095.afa --tree ureC_raxML.raxml.bestModel_rooted.tree --query ureC_10000_fragments_mafft_queriesonly.afa --model "Q.PFAM+R10{0.051755/0.163096/0.288424/0.431771/0.599218/0.800573/1.053245/1.393043/1.916291/3.302585}{0.129458/0.108723/0.121044/0.116191/0.104157/0.099554/0.107772/0.071831/0.070528/0.070742}" -T 12

however, it is only using the one set of parameters (weights) and seems to be re-optimising the second (rates). This is what I get in the .log

INFO Selected: Output dir: ./
INFO Selected: Query file: ureC_10000_fragments_mafft_queriesonly.afa
INFO Selected: Tree file: ureC_raxML.raxml.bestModel_rooted.tree
INFO Selected: Reference MSA: final_ureC_nogap095.afa
INFO Selected: Automatic switching of use of per rate scalers
INFO Selected: Preserving the root of the input tree
INFO Selected: Specified model: Q.PFAM+R10{0.051755/0.163096/0.288424/0.431771/0.599218/0.800573/1.053245/1.393043/1.916291/3.302585}{0.129458/0.108723/0.121044/0.116191/0.104157/0.099554/0.107772/0.071831/0.070528/0.070742}
INFO Selected: Using threads: 12

INFO Using model parameters:
INFO Rate heterogeneity: FREE (10 cats, mean), weights&rates: (0.129458,0.0620576) (0.108723,0.195563) (0.121044,0.345839) (0.116191,0.517722) (0.104157,0.718502) (0.099554,0.959939) (0.107772,1.26291) (0.071831,1.67035) (0.070528,2.29776) (0.070742,3.96002)

Thank you very much.

Best regards,

Alex

[lczech]

Hi Alex,

not sure if the model optimization can be deactivated. @pierrebarbera might have more insights into that.

So long
Lucas