-m MFP = default?

[BjoernStelbrink]

Dear all,

I am currently preparing a molecular phylogenetics course for our lab members, also including a tutorial on IQ-TREE 2. I came across the tip "Starting with version 1.5.4, -m MFP is the default behavior. Thus, this run is equivalent to iqtree -s example.phy" at https://iqtree.github.io/doc/Tutorial. I now noted, however, that both commands give different phylogenies and branch lengths, resulting from different best-fit models selected. So is that "tip" on the website not valid for IQ-TREE2?

Thank you and best wishes

Bjoern

[roblanf]

Hi Bjoern,

They should be equivalent in IQ-TREE2. It's possible that the differences in trees and branch lengths are due to stochastic effects on model selection and optimisation. There are two things you could do here:

1. Post the input and output files from both runs and I can have a look
2. Try running both with -seed 7 -T 1. This run everything on a single thread, with the same random number seed. This will ensure that identical analyses should give identical outputs

Yours,

Rob

[BjoernStelbrink]

Hi Rob,

thank you for your offer and further suggestions - this is very kind!

Find attached the output files

1. regular
2. +MFP
3. regular-seed (as suggested)
   4 +MFP-seed (as suggested)

The seed options produce the same results, but the tree topologies between standard commands vs. +MFP still differ. The same applies to the best-fit substitution models identified:

*. best_model.nex

###########

regular

###########

charpartition mymodels =
GTR{0.99709,18.027,0.589752,1.88391,32.0568}+F{0.360963,0.160571,0.195842,0.282624}: p1_12S_final_mafft{0.483023},
GTR{1.44277,17.8571,1.31882,1.3884,26.5604}+F{0.373015,0.148884,0.182608,0.295494}: p2_16S_final_mafft{0.578773},
GTR{2.20514,56.9591,0.167447,1.65655,38.9021}+F{0.212938,0.148965,0.194702,0.443394}: p3_ATP6_final{0.597541},
GTR{0.0001,44.3461,0.316497,0.0001,18.914}+F{0.267956,0.0861448,0.212626,0.433273}: p4_ATP8_final{0.512126},
GTR{3.32876,43.3717,1.50365,1.77299,43.5482}+F{0.242866,0.169056,0.200521,0.387556}: p5_COX1_final{0.799928},
GTR{1.23868,30.7543,0.565815,1.86888,31.5744}+F{0.264976,0.161817,0.207618,0.365589}: p6_COX2_final{0.523571},
GTR{3.17356,30.4643,0.348502,1.79245,33.125}+F{0.236546,0.231655,0.194237,0.337562}: p7_COX3_final{0.575153},
GTR{2.40091,30.2558,0.147722,2.19794,23.3011}+F{0.265163,0.217897,0.154398,0.362542}: p8_CYTB_final{0.678174},
GTR{1.25894,29.3083,0.2011,1.08764,24.7032}+F{0.25078,0.182526,0.178212,0.388482}: p9_ND1_final{0.603064},
GTR{4.66916,72.814,0.744335,5.96862,66.5983}+F{0.20759,0.112988,0.235509,0.443913}: p10_ND2_final{0.634509},
GTR{4.45667,58.0906,0.413936,2.35705,41.2326}+F{0.218442,0.140906,0.209457,0.431196}: p11_ND3_final{0.695237},
GTR{1.71804,31.4641,0.0001,4.46361,29.8216}+F{0.279876,0.165009,0.181734,0.373381}: p12_ND4L_final{0.478307},
GTR{2.63776,48.8338,0.123172,3.59095,32.9039}+F{0.269737,0.178231,0.150004,0.402028}: p13_ND4_final{0.689932},
GTR{2.57441,55.4081,0.322994,2.42783,40.3417}+F{0.272348,0.195201,0.159149,0.373302}: p14_ND5_final{0.696594},
GTR{1.12757,23.205,0.0001,1.16572,14.2836}+F{0.270601,0.192354,0.145787,0.391258}: p15_ND6_final{0.675213};
end;

#######

MFP

#######

charpartition mymodels =
TN{20.4899,36.3155}+F{0.360963,0.160571,0.195842,0.282624}+I{0.468712}+G4{0.661245}: p1_12S_final_mafft{0.527634},
TN{20.5522,31.0331}+F{0.373015,0.148884,0.182608,0.295494}+I{0.572404}+R2{0.380435,1.60376,0.0471608,8.26687}: p2_16S_final_mafft{1.0035},
HKY{54.3369}+F{0.212938,0.148965,0.194702,0.443394}+G4{0.223191}: p3_ATP6_final{0.655502},
HKY{61.0011}+F{0.267956,0.0861448,0.212626,0.433273}+I{0.49864}+R2{0.483769,1.50694,0.0175917,15.4045}: p4_ATP8_final{0.562072},
TPM3u{3.15333,61.8876}+F{0.242866,0.169056,0.200521,0.387556}+I{0.650689}+R2{0.305148,1.94476,0.0441628,9.20593}: p5_COX1_final{1.52577},
HKY{34.8211}+F{0.264976,0.161817,0.207618,0.365589}+G4{0.169582}: p6_COX2_final{0.561782},
TPM3u{3.8896,53.3693}+F{0.236546,0.231655,0.194237,0.337562}+I{0.574546}+R2{0.351089,1.39191,0.0743649,6.87577}: p7_COX3_final{0.622609},
TIM3{5.30661,82.0813,55.7234}+F{0.265163,0.217897,0.154398,0.362542}+I{0.566746}+R2{0.397801,1.71578,0.0354534,8.95437}: p8_CYTB_final{0.746608},
HKY{38.9733}+F{0.25078,0.182526,0.178212,0.388482}+I{0.559863}+R2{0.392492,1.57144,0.0476444,8.04339}: p9_ND1_final{0.647626},
TPM3u{7.39708,96.281}+F{0.20759,0.112988,0.235509,0.443913}+I{0.567175}+R2{0.383553,1.76135,0.0492725,6.5844}: p10_ND2_final{0.701834},
TPM3u{7.82429,105.98}+F{0.218442,0.140906,0.209457,0.431196}+I{0.587827}+R3{0.103875,1.02657,0.298109,2.37671,0.0101885,18.1426}: p11_ND3_final{0.799383},
TPM3u{12.9428,128.793}+F{0.279876,0.165009,0.181734,0.373381}+G4{0.301731}: p12_ND4L_final{0.524964},
TIM3{5.29821,100,62.7966}+F{0.269737,0.178231,0.150004,0.402028}+G4{0.268225}: p13_ND4_final{0.73856},
TIM3{4.23397,100,67.9951}+F{0.272348,0.195201,0.159149,0.373302}+I{0.390355}+G4{0.82391}: p14_ND5_final{0.774189},
TIM3{4.04125,100,51.3752}+F{0.270601,0.192354,0.145787,0.391258}+I{0.508066}+R2{0.441733,1.50315,0.0502017,6.6932}: p15_ND6_final{0.735426};
end;

################

regular-seed

################

charpartition mymodels =
GTR{0.997131,18.0123,0.589511,1.90104,32.0967}+F{0.360963,0.160571,0.195842,0.282624}: p1_12S_final_mafft{0.482983},
GTR{1.44329,17.8632,1.31932,1.38893,26.5706}+F{0.373015,0.148884,0.182608,0.295494}: p2_16S_final_mafft{0.578731},
GTR{2.23075,57.6264,0.169976,1.67498,39.3543}+F{0.212938,0.148965,0.194702,0.443394}: p3_ATP6_final{0.597503},
GTR{0.0001,31.9812,0.222607,0.0001,13.8351}+F{0.267956,0.0861448,0.212626,0.433273}: p4_ATP8_final{0.512411},
GTR{3.32834,43.3659,1.50335,1.77287,43.5439}+F{0.242866,0.169056,0.200521,0.387556}: p5_COX1_final{0.799875},
GTR{1.31576,32.6611,0.600614,1.97683,33.5375}+F{0.264976,0.161817,0.207618,0.365589}: p6_COX2_final{0.523569},
GTR{3.33329,31.9948,0.363582,1.88179,34.8299}+F{0.236546,0.231655,0.194237,0.337562}: p7_COX3_final{0.575125},
GTR{2.41892,30.45,0.148549,2.20855,23.4746}+F{0.265163,0.217897,0.154398,0.362542}: p8_CYTB_final{0.678138},
GTR{1.26238,29.3989,0.201787,1.09058,24.7364}+F{0.25078,0.182526,0.178212,0.388482}: p9_ND1_final{0.603028},
GTR{4.84142,75.534,0.784759,6.27908,69.0813}+F{0.20759,0.112988,0.235509,0.443913}: p10_ND2_final{0.634486},
GTR{4.3674,57.3102,0.407231,2.3282,40.7156}+F{0.218442,0.140906,0.209457,0.431196}: p11_ND3_final{0.695171},
GTR{1.72651,31.539,0.0001,4.61981,29.7483}+F{0.279876,0.165009,0.181734,0.373381}: p12_ND4L_final{0.47827},
GTR{2.65316,49.093,0.124058,3.59868,33.1269}+F{0.269737,0.178231,0.150004,0.402028}: p13_ND4_final{0.689886},
GTR{2.57529,55.4281,0.323203,2.42865,40.3557}+F{0.272348,0.195201,0.159149,0.373302}: p14_ND5_final{0.696544},
GTR{1.13571,23.214,0.0001,1.16688,14.3016}+F{0.270601,0.192354,0.145787,0.391258}: p15_ND6_final{0.675135};
end;

############

MFP-seed

############

charpartition mymodels =
TN{20.4851,36.3024}+F{0.360963,0.160571,0.195842,0.282624}+I{0.464541}+G4{0.649855}: p1_12S_final_mafft{0.527242},
TN{20.5521,31.0277}+F{0.373015,0.148884,0.182608,0.295494}+I{0.572511}+R2{0.380433,1.60512,0.0470559,8.27442}: p2_16S_final_mafft{1.00345},
HKY{54.3325}+F{0.212938,0.148965,0.194702,0.443394}+G4{0.223227}: p3_ATP6_final{0.655234},
HKY{60.9932}+F{0.267956,0.0861448,0.212626,0.433273}+I{0.498773}+R2{0.483642,1.50735,0.017585,15.4097}: p4_ATP8_final{0.5622},
TPM3u{3.14566,61.7736}+F{0.242866,0.169056,0.200521,0.387556}+I{0.650724}+R2{0.305135,1.94545,0.0441408,9.20635}: p5_COX1_final{1.52553},
HKY{34.8197}+F{0.264976,0.161817,0.207618,0.365589}+G4{0.169431}: p6_COX2_final{0.561558},
TPM3u{3.89024,53.3793}+F{0.236546,0.231655,0.194237,0.337562}+I{0.574822}+R2{0.350953,1.39413,0.0742252,6.88076}: p7_COX3_final{0.622671},
TIM3{5.30285,82.029,55.6856}+F{0.265163,0.217897,0.154398,0.362542}+I{0.56685}+R2{0.397747,1.71672,0.0354028,8.95913}: p8_CYTB_final{0.746689},
HKY{38.9661}+F{0.25078,0.182526,0.178212,0.388482}+I{0.560003}+R2{0.392437,1.57264,0.0475602,8.04959}: p9_ND1_final{0.647683},
TPM3u{7.39842,96.2984}+F{0.20759,0.112988,0.235509,0.443913}+I{0.567078}+R2{0.383532,1.76007,0.0493898,6.57941}: p10_ND2_final{0.701796},
TPM3u{7.2543,100}+F{0.218442,0.140906,0.209457,0.431196}+I{0.605715}+R2{0.383231,2.10594,0.0110544,17.4538}: p11_ND3_final{0.796754},
HKY{26.9161}+F{0.279876,0.165009,0.181734,0.373381}+G4{0.297115}: p12_ND4L_final{0.522553},
TIM3{5.29849,100,62.7959}+F{0.269737,0.178231,0.150004,0.402028}+G4{0.26827}: p13_ND4_final{0.738266},
TIM3{4.23529,100,68.0223}+F{0.272348,0.195201,0.159149,0.373302}+I{0.382427}+G4{0.797355}: p14_ND5_final{0.774274},
TIM3{4.04178,100,51.3819}+F{0.270601,0.192354,0.145787,0.391258}+I{0.507997}+R2{0.441756,1.50266,0.0502469,6.69082}: p15_ND6_final{0.73538};
end;

But maybe I missed something

Cheers

Bjoern
MFP_seed.zip
regular_seed.zip
MFP.zip
regular.zip

[roblanf]

HI Bjoern,

So it certainly looks like the defaults have changed here. @thomaskf can you take a look at this? @bqminh was this the intention? My feeling here is that having -MFP be the default is very sensible, because it represents best practice for most people.

In this case though, running without explicitly calling -MFP puts a GTR model on every partition (unexpected). Running with -MFP does model selection on every partition (as expected).

@BjoernStelbrink - it might take us until February to get this properly straightened out (it's the long summer break in Australia now), but for your workshop I suggest just explicitly calling -MFP for now.

Thanks for bringing this to our attention!

Rob

[BjoernStelbrink]

Hi Rob,

thanks for your reply. I like the flexibility of the various commands in IQ-TREE 2, and it is probably common practice for users to employ -MFP rather than using the default settings. I did not expect much difference between an established model such as GTR+G and the ones selected by MF, but I noticed too many topological differences in that case.

Thank you for your support and happy holidays

Bjoern

[roblanf]

Agreed that differences are usually relatively rare and/or occur on very short or poorly supported branches. There are a couple of papers on this, so if you're seeing important differences (e.g. on long well supported branches) with model selection, you have an interesting case study. It can happen though, here's a nice paper on the subject:

https://shenxlab.com/pdfs/2017_Shen_etal_NEE.pdf