Re: [iqtree] partition specific rates - evol. rates

61 views
Skip to first unread message

Minh Bui

unread,
Mar 1, 2021, 9:20:53 PM3/1/21
to IQ-TREE
Hi Karen,

Yes they are exactly the same, might look different due to rounding.

Minh

On 28 Feb 2021, at 12:11 am, 'kmeus...@googlemail.com' via IQ-TREE <iqt...@googlegroups.com> wrote:

Dear all,

just a quick question (using v.1.6.12)

the difference between the "partition specific rates" in the *.log file and the "speed" in the *.iqtree file is the same correct (only differes according to rounding the numers) ? or I I would like to resport the "evolut. rates" should I stick to the "speed" value - and below is still valid?

(from a post 3 years ago: "Speed column is exactly as you asked: it is the partition specific rate. They are computed by maximum likelihood. The rates are normalized such that the weighted average is 1, where the weights are the partition alignment length divided by supermatrix length")

best Karen

--
You received this message because you are subscribed to the Google Groups "IQ-TREE" group.
To unsubscribe from this group and stop receiving emails from it, send an email to iqtree+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/iqtree/bc434479-b0a1-448e-82b7-d1738ba10baan%40googlegroups.com.

c.maye...@uni-bonn.de

unread,
Jul 23, 2021, 8:37:29 PM7/23/21
to IQ-TREE
Dear Minh,

as Karen I was wondering whether the input order of the partitions is the same as the order of the numbers in the "partition specific rates" (*.log) output
and whether speed (*.iqtree file) is the same as the "partition specific rates"?

After finding the answer, I checked my data and I could not find the correspondence that I would have expected:

For my data set the *.log file contains the following partition specific rates:
Partition-specific rates:  0.151 0.141 0.236 0.366 0.390 0.381 0.225 0.213 0.517 ....

In the corresponding *.iqtree file I find:
Edge-linked-proportional partition model with separate substitution models and separate rates across sites

  ID  Model           Speed  Parameters
   1  LG+R3          0.1554  LG+R3{0.823196,0.347324,0.162733,3.17633,0.0140707,14.0142}
   2  VT+R3          0.0880  VT+R3{0.888953,0.216238,0.089011,3.39085,0.0220356,22.9607}
   3  JTT+R3         0.1972  JTT+R3{0.841837,0.286363,0.133907,3.23114,0.024256,13.4505}
   4  JTT+R3         0.3208  JTT+R3{0.682353,0.204336,0.264993,1.82026,0.0526532,7.18312}
   5  JTT+R3         0.2231  JTT+R3{0.764902,0.0547832,0.206728,2.21628,0.0283696,17.622}
   6  LG+R3          0.4109  LG+R3{0.646799,0.171039,0.328276,2.08465,0.0249248,8.22599}
   7  JTT+R4         0.2162  JTT+R4{0.682751,0.151679,0.241186,1.58744,0.0697556,5.37925,0.00630789,21.9312}
   8  JTTDCMut+R4    0.2303  JTTDCMut+R4{0.423401,0.0336106,0.398709,0.847233,0.164368,3.20274,0.0135223,8.98831}
   9  LG+R4          0.5182  LG+R4{0.337071,0.0207592,0.399248,0.551705,0.221075,2.15266,0.0426058,6.96706}

You wrote that they are exactly the same but might look different due to rounding.
But I do not have this correspondence between the numbers in the rates line and the speed column (3rd column in the *.iqtree file).
The difference cannot be explained by rounding differences, nor by re weighting the numbers.

Well, it looks as if:
- they are not in the same order?
- they are different things?

Many thanks for your help in advance.

Christoph

kmeus...@googlemail.com

unread,
Aug 8, 2021, 10:20:45 AM8/8/21
to IQ-TREE
Dear Minh,

can you pls answer this question?
I looked up again my files: in my files values were the same. Could the difference be due to that I used 1.6.X and Chrsitoph the new version 2.X?

Best, Karen

Minh Bui

unread,
Aug 9, 2021, 2:31:14 AM8/9/21
to IQ-TREE, c.maye...@uni-bonn.de
Hi Christoph and Karen,

Sorry I missed this email.

Am I right that .log and .iqtree files are produced by one run of IQ-TREE? If yes, the last line with “partition-specific rates” from the log file (I mean the last line, because this line might be printed several times during the progress of the analysis) and the “Speed” column from the iqtree report should be the same, just printed with different precision. And the order should be the same as when loading the partitions.

Minh

Christoph Mayer

unread,
Aug 10, 2021, 8:17:59 PM8/10/21
to Minh Bui, IQ-TREE
Dear Minh,

many thanks for your answer! Indeed the rates are printed multiple times and I looked at the first occurrence and not the last occurrence.
The numbers in the last occurring "Partition-specific rates:“ line are indeed almost equal to the speed values in the .iqtree file.

One more thing confused me: the order of the partitions is not always the same in the .iqtree and the .log file.

Within the .iqtree file the there are two tables with the first column header „ID“. Furthermore there is the „Best-fit model according to BIC line.
In all three, the order seems to be the same.

In the .log file, there is a table starting with the lines:
NOTE: ModelFinder requires 15850 MB RAM!
Selecting individual models for 2413 charsets using BIC...
 No. Model        Score       TreeLen     Charset“

In this table the order is different. This confused me at the beginning and led me invest some time to double check the orders in the .iqtree file. 

Anyway. Many thanks again for your answer and for the great software!!!

Best
Christoph

****************************************
Dr. Christoph Mayer
Email:  c.maye...@uni-bonn.de
Tel.:   +49 (0)228 9122 403

Zoologisches Forschungsmuseum Alexander Koenig
- Leibniz Institut für Biodiversität der Tiere -
Adenauerallee 160
53113 Bonn, Germany 
www.zfmk.de

Stiftung des öffentlichen Rechts;
Direktor: Prof. B. Misof
Sitz: Bonn
****************************************





Minh Bui

unread,
Aug 11, 2021, 7:26:37 PM8/11/21
to c.maye...@uni-bonn.de, IQ-TREE
Hi Christoph,

Yes, this model selection step has a different order of printing partitions for a reason. In a multi-threading run, IQ-TREE will use each thread to select best model for each partition. To acheive high parallel efficiency, it will sort the partitions in descending order of partition sizes, such that large partitions will be processed first before small partitions. Therefore, the first partition you see is the one finished first among the threads. And so on. So there is no particular order, it may change with different runs with different number of threads. 

Hope that helps,
Minh
Reply all
Reply to author
Forward
0 new messages