Model specified gets ignored

17 views

Skip to first unread message

Émilie Tremblay

unread,

Jan 16, 2025, 10:48:47 PMJan 16

to raxml

Hello, I know this has been reported before, but I tried what has been suggested before without any luck.

My computer has 56 cpus, 128GB

After using modeltest-ng to determine which model to use in raxml-ng, it indicated to pick the GTR+G4 model.

I have version 1.2.2 of raxml-ng and this is the command I run:

raxml-ng --msa /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta --msa-format FASTA --data-type DNA --seed 1234 -model GTR+G4 --threads 48

this outputs the following warning:
WARNING: The model you specified on the command line (GTR+G4) will be ignored
since the binary MSA file already contains a model definition.
If you want to change the model, please re-run RAxML-NG
with the original PHYLIP/FASTA alignment and --redo option.

Can you please help me sort this out?

My alignment is a bunch of fasta files that are very similar (ITS1 fungal sequences) of about 230bp in average.

By the way, what the modeltest-ng run indicated for me to run seemed pretty limited in options. Is this concerning ? I would have like to know more on bootrapping for instance.

modeltest-ng -i /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta -d nt -p 48 -T raxml -r 1234 -o /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/model-test-output_fungi

output:

Input data:
MSA: /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta
Tree: Maximum parsimony
file: -
#taxa: 3152
#sites: 2283
#patterns: 1917
Max. thread mem: 2995 MB

Output:
Log: /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/model-test-output_fungi.log
Starting tree: /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/model-test-output_fungi.tree
Results: /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/model-test-output_fungi.out

Selection options:
# dna schemes: 3
# dna models: 12
include model parameters:
Uniform: true
p-inv (+I): true
gamma (+G): true
both (+I+G): true
free rates (+R): false
fixed freqs: false
estimated freqs: true
#categories: 4
gamma rates mode: mean
asc bias: none
epsilon (opt): 0.01
epsilon (par): 0.05
keep branches: false

Additional options:
verbosity: very low
threads: 48/28
RNG seed: 1234
subtree repeats: enabled
--------------------------------------------------------------------------------

BIC model K lnL score delta weight
--------------------------------------------------------------------------------
1 GTR+G4 9 -360262.0260 769320.8321 0.0000 1.0000
2 GTR+I+G4 10 -360286.8997 769378.3126 57.4805 0.0000
3 GTR 8 -385567.0516 819923.1500 50602.3179 0.0000
4 GTR+I 9 -385575.1867 819947.1535 50626.3214 0.0000
5 HKY 4 -386346.6207 821451.3552 52130.5231 0.0000
6 HKY+I 5 -386351.7237 821469.2944 52148.4623 0.0000
7 F81 3 -387996.7704 824743.9214 55423.0892 0.0000
8 F81+I 4 -388005.0147 824768.1432 55447.3111 0.0000
9 HKY+G4 5 -429866.2668 908498.3807 139177.5486 0.0000
10 HKY+I+G4 6 -429976.1962 908725.9726 139405.1405 0.0000

Best model according to BIC
---------------------------
Model: GTR+G4
lnL: -360262.0260
Frequencies: 0.2585 0.2369 0.2170 0.2877
Subst. Rates: 1.2536 1.7666 1.6104 0.9823 2.4136 1.0000
Inv. sites prop: -
Gamma shape: 0.6464
Score: 769320.8321
Weight: 1.0000
---------------------------
Parameter importances
---------------------------
P.Inv: -
Gamma: 1.0000
Gamma-Inv: 0.0000
Frequencies: 1.0000
---------------------------
Model averaged estimates
---------------------------
P.Inv: -
Alpha: 0.6464
Alpha-P.Inv: 0.6472
P.Inv-Alpha: 0.0221
Frequencies: 0.2585 0.2369 0.2170 0.2877

Commands:
> phyml -i /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta -m 012345 -f m -v 0 -a e -c 4 -o tlr
> raxmlHPC-SSE3 -s /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta -m GTRGAMMAX -n EXEC_NAME -p PA>
> raxml-ng --msa /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta --model GTR+G4
> paup -s /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta
> iqtree -s /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta -m GTR+G4

AIC model K lnL score delta weight
--------------------------------------------------------------------------------
1 GTR+G4 9 -360262.0260 733144.0521 0.0000 1.0000
2 GTR+I+G4 10 -360286.8997 733195.7993 51.7473 0.0000
3 GTR 8 -385567.0516 783752.1032 50608.0511 0.0000
4 GTR+I 9 -385575.1867 783770.3735 50626.3214 0.0000
5 HKY 4 -386346.6207 785303.2414 52159.1893 0.0000
6 HKY+I 5 -386351.7237 785315.4474 52171.3953 0.0000
7 F81 3 -387996.7704 788601.5408 55457.4887 0.0000
8 F81+I 4 -388005.0147 788620.0294 55475.9773 0.0000
9 HKY+G4 5 -429866.2668 872344.5337 139200.4816 0.0000
10 HKY+I+G4 6 -429976.1962 872566.3923 139422.3402 0.0000
--------------------------------------------------------------------------------
Best model according to AIC

Best model according to AIC
---------------------------
Model: GTR+G4
lnL: -360262.0260
Frequencies: 0.2585 0.2369 0.2170 0.2877
Subst. Rates: 1.2536 1.7666 1.6104 0.9823 2.4136 1.0000
Inv. sites prop: -
Gamma shape: 0.6464
Score: 733144.0521
Weight: 1.0000
---------------------------
Parameter importances
---------------------------
P.Inv: -
Gamma: 1.0000
Gamma-Inv: 0.0000
Frequencies: 1.0000
---------------------------
Model averaged estimates
---------------------------
P.Inv: -
Alpha: 0.6464
Alpha-P.Inv: 0.6472
P.Inv-Alpha: 0.0221
Frequencies: 0.2585 0.2369 0.2170 0.2877

Commands:
> phyml -i /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta -m 012345 -f m -v 0 -a e -c 4 -o tlr
> raxmlHPC-SSE3 -s /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta -m GTRGAMMAX -n EXEC_NAME -p PA>
> raxml-ng --msa /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta --model GTR+G4
> paup -s /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta
> iqtree -s /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta -m GTR+G4

AICc model K lnL score delta weight
--------------------------------------------------------------------------------
1 F81 3 -387996.7704 80282041.5408 0.0000 1.0000
2 HKY 4 -386346.6207 80303963.2414 21921.7006 0.0000
3 F81+I 4 -388005.0147 80307280.0294 25238.4886 0.0000
4 HKY+I 5 -386351.7237 80329199.4474 47157.9066 0.0000
5 GTR+G4 9 -360262.0260 80377964.0521 95922.5113 0.0000
6 F81+G4 4 -431215.1159 80393700.2318 111658.6910 0.0000
7 GTR+I+G4 10 -360286.8997 80403259.7993 121218.2585 0.0000
8 GTR 8 -385567.0516 80403332.1032 121290.5624 0.0000
9 HKY+G4 5 -429866.2668 80416228.5337 134186.9929 0.0000
10 F81+I+G4 5 -431325.7187 80419147.4374 137105.8966 0.0000

Best model according to AICc
---------------------------
Model: F81
lnL: -387996.7704
Frequencies: 0.2702 0.2471 0.1822 0.3005
Subst. Rates: 1.0000 1.0000 1.0000 1.0000 1.0000 1.0000
Inv. sites prop: -
Gamma shape: -
Score: 80282041.5408
Weight: 1.0000
---------------------------
Parameter importances
---------------------------
P.Inv: -
Gamma: -
Gamma-Inv: -
Frequencies: 1.0000
---------------------------
Model averaged estimates
---------------------------
P.Inv: -
Alpha: -
Alpha-P.Inv: -
P.Inv-Alpha: -
Frequencies: 0.2702 0.2471 0.1822 0.3005

Commands:
> phyml -i /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta -m 000000 -f m -v 0 -a 0 -c 1 -o tlr
> raxmlHPC-SSE3 -s /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta -c 1 -m GTRCATX --JC69 -n EXEC_>
> raxml-ng --msa /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta --model F81
> paup -s /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta
> iqtree -s /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/ASV_seqs_MSA.fasta -m F81

Done

Thanks

Grimm

unread,

Jan 17, 2025, 7:54:04 AMJan 17

to raxml

Hi,

the error message is quite clear, the ASV_seqs_MSA.fasta either contains already a model-definition or could have the wrong data format (e.g. binary instead of nucleotide sequence alignment).

Ideally, it should look like this:

>ASV1

GCATTCCCTTTTRG....

>ASV2

GCATCCCCTATARG....

ASV1 GCATTCCCTTTTRG....

ASV2 GCATCCCCTATARG....

and nothing else. Just labels and sequences.

Re. Second question. Nah, that's fine. For non-coding but functional spacers that underly few structural constraints as in the case of ITS1 (in contrast to ITS2, the ITS1 is not folded during rRNA maturation) but are also not 100% free to evolve neutrally in samples that comprise a high diversity (1900 DAP for 2200 AS; I guess it's an environmental bulk sample), modeltesting usually converges to a single optimal option, wether max-parametrised GTR (typical result for most organism's ITS1) or classic F81 (next-most, together with HKY) depends on the sequence structure (mutation patterns in globally aligned sites) and stuff like overall GC-content.

Cheers, G.

Oleksiy Kozlov

unread,

Jan 17, 2025, 7:55:05 AMJan 17

to ra...@googlegroups.com

Hi Émilie,

> raxml-ng --msa /home/tremblayemi/nfcore/rawData/results/fungi-and-controls_v2/dada2/

> ASV_seqs_MSA.fasta --msa-format FASTA --data-type DNA --seed 1234 -model GTR+G4 --threads 48
>
> this outputs the following warning:
> WARNING: The model you specified on the command line (GTR+G4) will be ignored
> since the binary MSA file already contains a model definition.
> If you want to change the model, please re-run RAxML-NG
> with the original PHYLIP/FASTA alignment and --redo option.
>
> Can you please help me sort this out?

Please ignore this warning since model is the same (GTR+G4) in both cases.

> By the way, what the modeltest-ng run indicated for me to run seemed pretty limited in options. Is
> this concerning ? I would have like to know more on bootrapping for instance.

Not quite sure what do you mean by "limited" here?
Bootstrapping is typically used during tree search with raxml-ng, and not for model selection.

Best,
Oleksiy

Reply all

Reply to author

Forward

0 new messages