From b9db03cca99524b7c0db191f4689f74262d45850 Mon Sep 17 00:00:00 2001
From: HuaiyanRen <928003585@qq.com>
Date: Tue, 11 Aug 2026 15:03:48 +1000
Subject: [PATCH 01/10] Update PartitionFinder-mAIC
---
doc/Advanced-Tutorial.md | 20 ++++++++++----------
1 file changed, 10 insertions(+), 10 deletions(-)
diff --git a/doc/Advanced-Tutorial.md b/doc/Advanced-Tutorial.md
index d64b32c..c9bf88f 100644
--- a/doc/Advanced-Tutorial.md
+++ b/doc/Advanced-Tutorial.md
@@ -143,8 +143,7 @@ Choosing the right partitioning scheme
--------------------------------------
-ModelFinder implements a greedy strategy ([Lanfear et al., 2012]) that starts with the full partition model and subsequentially
-merges two genes until the model fit does not increase any further:
+ModelFinder integrates the PartitionFinder algorithm ([Lanfear et al., 2012]) that starts with the full partition model and subsequently merges two partition subsets until the model fit does not increase any further:
iqtree -s example.phy -p example.nex -m MFP+MERGE
# for version 1.x change -p to -spp
@@ -166,16 +165,16 @@ To resemble PartitionFinder and save time:
iqtree -s example.phy -p example.nex -m TESTMERGEONLY
# for version 1.x change -p to -spp
+This implementation includes a greedy algorithm ([Lanfear et al., 2012]) that subsequentially merges a single pair of subsets per iteration, and a *relaxed hierarchical clustering algorithm* (default; [Lanfear et al., 2014]) that merges multiple pairs of subsets per iteration to reduce the computational burden. Starting with version 3.1.3, PartitionFinder by default applies marginal Akaike Information Criterion (mAIC; [Susko et al., 2026]) as the criterion for partition merging. If you use PartitionFinder-mAIC in a publication, please cite:
-To reduce the computational burden IQ-TREE implements the *relaxed hierarchical clustering algorithm* ([Lanfear et al., 2014]), which is invoked via `-rcluster` option:
-
- iqtree -s example.phy -p example.nex -m MF+MERGE -rcluster 10
- # for version 1.x change -p to -spp
-
-
-to only examine the top 10% partition merging schemes (similar to the `--rcluster-percent 10` option in PartitionFinder).
-
+> TBD
+Here are the options to specify the detail of PartitionFinder algorithm:
+| Option | Description |
+| ------------ | ---------------------------------------------------------------------- |
+| `--merge` | Specify either `rcluster` or `greedy` algorithm. *DEFAULT: `rcluster`* |
+| `--rcluster` | Specify the percentage of top partition merging schemes. *DEFAULT: 10* |
+| `-merit` | Specify either `mAIC`, `AIC`,`AICc` or `BIC` for partition merging criterion. *DEFAULT: `mAIC`* |
Ultrafast bootstrapping with partition model
--------------------------------------------
@@ -587,6 +586,7 @@ See [Command Reference](Command-Reference) for a complete list of all options av
[Kishino and Hasegawa, 1989]: https://doi.org/10.1007/BF02100115
[Lanfear et al., 2012]: https://doi.org/10.1093/molbev/mss020
[Lanfear et al., 2014]: https://doi.org/10.1186/1471-2148-14-82
+[Susko et al., 2026]: https://doi.org/10.1093/sysbio/syag013
[Lopez et al., 2002]: http://mbe.oxfordjournals.org/content/19/1/1.full
[Nei et al., 2001]: https://doi.org/10.1073/pnas.051611498
[Seo et al., 2005]: https://doi.org/10.1073/pnas.0408313102
From 711bf16c26ffc01231919143ec978364697ad884 Mon Sep 17 00:00:00 2001
From: HuaiyanRen <928003585@qq.com>
Date: Tue, 11 Aug 2026 15:12:08 +1000
Subject: [PATCH 02/10] Update Advanced-Tutorial.md
---
doc/Advanced-Tutorial.md | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/doc/Advanced-Tutorial.md b/doc/Advanced-Tutorial.md
index c9bf88f..2159723 100644
--- a/doc/Advanced-Tutorial.md
+++ b/doc/Advanced-Tutorial.md
@@ -173,7 +173,7 @@ Here are the options to specify the detail of PartitionFinder algorithm:
| Option | Description |
| ------------ | ---------------------------------------------------------------------- |
| `--merge` | Specify either `rcluster` or `greedy` algorithm. *DEFAULT: `rcluster`* |
-| `--rcluster` | Specify the percentage of top partition merging schemes. *DEFAULT: 10* |
+| `--rcluster` | Specify the percentage of most similar candidate partition pairs retained for merging at each iteration (relaxed clustering algorithm only). *DEFAULT: 10* |
| `-merit` | Specify either `mAIC`, `AIC`,`AICc` or `BIC` for partition merging criterion. *DEFAULT: `mAIC`* |
Ultrafast bootstrapping with partition model
From 9ae29ad39b85b5dfb70e2ad9cea62cc574df6051 Mon Sep 17 00:00:00 2001
From: HuaiyanRen <928003585@qq.com>
Date: Tue, 11 Aug 2026 15:13:51 +1000
Subject: [PATCH 03/10] Update Advanced-Tutorial.md
---
doc/Advanced-Tutorial.md | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/doc/Advanced-Tutorial.md b/doc/Advanced-Tutorial.md
index 2159723..c70b2a8 100644
--- a/doc/Advanced-Tutorial.md
+++ b/doc/Advanced-Tutorial.md
@@ -172,9 +172,10 @@ This implementation includes a greedy algorithm ([Lanfear et al., 2012]) that su
Here are the options to specify the detail of PartitionFinder algorithm:
| Option | Description |
| ------------ | ---------------------------------------------------------------------- |
+| `-merit` | Specify either `mAIC`, `AIC`,`AICc` or `BIC` for partition merging criterion. *DEFAULT: `mAIC`* |
| `--merge` | Specify either `rcluster` or `greedy` algorithm. *DEFAULT: `rcluster`* |
| `--rcluster` | Specify the percentage of most similar candidate partition pairs retained for merging at each iteration (relaxed clustering algorithm only). *DEFAULT: 10* |
-| `-merit` | Specify either `mAIC`, `AIC`,`AICc` or `BIC` for partition merging criterion. *DEFAULT: `mAIC`* |
+
Ultrafast bootstrapping with partition model
--------------------------------------------
From 3cf036d90e2e42e921c99b63b015ba0c999017e5 Mon Sep 17 00:00:00 2001
From: HuaiyanRen <928003585@qq.com>
Date: Wed, 19 Aug 2026 14:45:46 +1000
Subject: [PATCH 04/10] Update PartitionFinder details
---
doc/Advanced-Tutorial.md | 29 ++++++++++++++++++++---------
1 file changed, 20 insertions(+), 9 deletions(-)
diff --git a/doc/Advanced-Tutorial.md b/doc/Advanced-Tutorial.md
index c70b2a8..6cb95d7 100644
--- a/doc/Advanced-Tutorial.md
+++ b/doc/Advanced-Tutorial.md
@@ -165,17 +165,27 @@ To resemble PartitionFinder and save time:
iqtree -s example.phy -p example.nex -m TESTMERGEONLY
# for version 1.x change -p to -spp
-This implementation includes a greedy algorithm ([Lanfear et al., 2012]) that subsequentially merges a single pair of subsets per iteration, and a *relaxed hierarchical clustering algorithm* (default; [Lanfear et al., 2014]) that merges multiple pairs of subsets per iteration to reduce the computational burden. Starting with version 3.1.3, PartitionFinder by default applies marginal Akaike Information Criterion (mAIC; [Susko et al., 2026]) as the criterion for partition merging. If you use PartitionFinder-mAIC in a publication, please cite:
-
+Starting with version 3.1.3, PartitionFinder by default applies the marginal Akaike Information Criterion (mAIC; [Susko et al., 2026]) as the criterion for partition merging. If you use PartitionFinder-mAIC in a publication, please cite:
> TBD
-Here are the options to specify the detail of PartitionFinder algorithm:
-| Option | Description |
-| ------------ | ---------------------------------------------------------------------- |
-| `-merit` | Specify either `mAIC`, `AIC`,`AICc` or `BIC` for partition merging criterion. *DEFAULT: `mAIC`* |
-| `--merge` | Specify either `rcluster` or `greedy` algorithm. *DEFAULT: `rcluster`* |
-| `--rcluster` | Specify the percentage of most similar candidate partition pairs retained for merging at each iteration (relaxed clustering algorithm only). *DEFAULT: 10* |
+PartitionFinder implementation includes four merging algorithms, which differ in how thoroughly they search the space of partitioning schemes and how many pairs they merge per iteration:
+
+| Algorithm | IQ-TREE Command | Description | Reference |
+| ------------------------------------ | --------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------- |
+| Greedy | `greedy` | At each iteration, evaluates all candidate subset pairs and merges the single pair that gives the best information-criterion score. | [Lanfear et al., 2012] |
+| Relaxed hierarchical clustering | `rcluster` | At each iteration, evaluates only the top k% most similar candidate subset pairs and merges the best-scoring one. | [Lanfear et al., 2014] |
+| Fast hierarchical relaxed clustering | `rclusterf` | At each iteration, evaluates only the top k% most similar candidate subset pairs and merges multiple compatible pairs at once. (IQ-TREE default). | [Lanfear et al., 2014] |
+| k-means | `kmeans` | Estimates an evolutionary rate for each site, then iteratively clusters individual sites by rate using k-means, without relying on predefined data blocks. | [Frandsen et al., 2015] |
+
+The following options control the details of the merging procedure:
+| Option | Description |
+| ---------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
+| `-merit` | Specify either `mAIC`, `AIC`, `AICc` or `BIC` for the partition merging criterion. *DEFAULT: `mAIC`* |
+| `--merge` | Specify either `rclusterf`, `rcluster`, `greedy` or `kmeans` algorithm. *DEFAULT: `rclusterf`* |
+| `--rclusterf` | Specify the percentage of most similar candidate partition pairs retained for merging at each iteration (this automatically selects the fast relaxed clustering algorithm). *DEFAULT: 10* |
+| `--rcluster` | Specify the percentage of most similar candidate partition pairs retained for merging at each iteration (this automatically selects the relaxed clustering algorithm). *DEFAULT: 10* |
+| `--rcluster-max` | Specify the maximum number of most similar candidate partition pairs retained for merging at each iteration (only for the fast relaxed clustering and relaxed clustering algorithms). *DEFAULT: 10 × number of partitions in the full partition model* |
Ultrafast bootstrapping with partition model
--------------------------------------------
@@ -586,8 +596,9 @@ See [Command Reference](Command-Reference) for a complete list of all options av
[Kishino et al., 1990]: https://doi.org/10.1007/BF02109483
[Kishino and Hasegawa, 1989]: https://doi.org/10.1007/BF02100115
[Lanfear et al., 2012]: https://doi.org/10.1093/molbev/mss020
-[Lanfear et al., 2014]: https://doi.org/10.1186/1471-2148-14-82
[Susko et al., 2026]: https://doi.org/10.1093/sysbio/syag013
+[Lanfear et al., 2014]: https://doi.org/10.1186/1471-2148-14-82
+[Frandsen et al., 2015]: https://doi.org/10.1186/s12862-015-0283-7
[Lopez et al., 2002]: http://mbe.oxfordjournals.org/content/19/1/1.full
[Nei et al., 2001]: https://doi.org/10.1073/pnas.051611498
[Seo et al., 2005]: https://doi.org/10.1073/pnas.0408313102
From cba6bdad3e16634bf109e4b15e223a354429d540 Mon Sep 17 00:00:00 2001
From: HuaiyanRen <928003585@qq.com>
Date: Wed, 19 Aug 2026 15:00:40 +1000
Subject: [PATCH 05/10] Update PartitionFinder
---
doc/Advanced-Tutorial.md | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/doc/Advanced-Tutorial.md b/doc/Advanced-Tutorial.md
index 6cb95d7..973b5f0 100644
--- a/doc/Advanced-Tutorial.md
+++ b/doc/Advanced-Tutorial.md
@@ -181,7 +181,7 @@ The following options control the details of the merging procedure:
| Option | Description |
| ---------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
-| `-merit` | Specify either `mAIC`, `AIC`, `AICc` or `BIC` for the partition merging criterion. *DEFAULT: `mAIC`* |
+| `-merit` | Specify either `mAIC`, `AIC`, `AICc` or `BIC` for the partition merging criterion. *Defaults to `mAIC` when `MERGE` is enabled; defaults to `BIC` for stardard model selection.* |
| `--merge` | Specify either `rclusterf`, `rcluster`, `greedy` or `kmeans` algorithm. *DEFAULT: `rclusterf`* |
| `--rclusterf` | Specify the percentage of most similar candidate partition pairs retained for merging at each iteration (this automatically selects the fast relaxed clustering algorithm). *DEFAULT: 10* |
| `--rcluster` | Specify the percentage of most similar candidate partition pairs retained for merging at each iteration (this automatically selects the relaxed clustering algorithm). *DEFAULT: 10* |
From a8404ac875791d1dc057ea9091b9e2852c7e4f06 Mon Sep 17 00:00:00 2001
From: HuaiyanRen <928003585@qq.com>
Date: Wed, 26 Aug 2026 11:12:06 +1000
Subject: [PATCH 06/10] Set BIC default for PartitionFinder
---
doc/Advanced-Tutorial.md | 8 ++++++--
1 file changed, 6 insertions(+), 2 deletions(-)
diff --git a/doc/Advanced-Tutorial.md b/doc/Advanced-Tutorial.md
index 973b5f0..59d1e97 100644
--- a/doc/Advanced-Tutorial.md
+++ b/doc/Advanced-Tutorial.md
@@ -165,7 +165,11 @@ To resemble PartitionFinder and save time:
iqtree -s example.phy -p example.nex -m TESTMERGEONLY
# for version 1.x change -p to -spp
-Starting with version 3.1.3, PartitionFinder by default applies the marginal Akaike Information Criterion (mAIC; [Susko et al., 2026]) as the criterion for partition merging. If you use PartitionFinder-mAIC in a publication, please cite:
+Starting with version 3.1.4, PartitionFinder supports the marginal Akaike Information Criterion (mAIC; [Susko et al., 2026]) as an optional criterion for partition merging. To apply PartitionFinder-mAIC:
+
+ iqtree -s example.phy -p example.nex -m MFP+MERGE -merit mAIC
+
+If you use PartitionFinder-mAIC in a publication, please cite:
> TBD
PartitionFinder implementation includes four merging algorithms, which differ in how thoroughly they search the space of partitioning schemes and how many pairs they merge per iteration:
@@ -181,7 +185,7 @@ The following options control the details of the merging procedure:
| Option | Description |
| ---------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
-| `-merit` | Specify either `mAIC`, `AIC`, `AICc` or `BIC` for the partition merging criterion. *Defaults to `mAIC` when `MERGE` is enabled; defaults to `BIC` for stardard model selection.* |
+| `-merit` | Specify either `AIC`, `AICc`, `BIC` or `mAIC` for the partition merging criterion. *DEFAULT: `BIC`* |
| `--merge` | Specify either `rclusterf`, `rcluster`, `greedy` or `kmeans` algorithm. *DEFAULT: `rclusterf`* |
| `--rclusterf` | Specify the percentage of most similar candidate partition pairs retained for merging at each iteration (this automatically selects the fast relaxed clustering algorithm). *DEFAULT: 10* |
| `--rcluster` | Specify the percentage of most similar candidate partition pairs retained for merging at each iteration (this automatically selects the relaxed clustering algorithm). *DEFAULT: 10* |
From 438457cfbbc6747697118ddcc3ae6c37a0b8948d Mon Sep 17 00:00:00 2001
From: HuaiyanRen <928003585@qq.com>
Date: Fri, 4 Sep 2026 19:16:34 +1000
Subject: [PATCH 07/10] Update PartitionFinder options
---
doc/Advanced-Tutorial.md | 19 +--------------
doc/Command-Reference.md | 51 ++++++++++++++++++++--------------------
2 files changed, 27 insertions(+), 43 deletions(-)
diff --git a/doc/Advanced-Tutorial.md b/doc/Advanced-Tutorial.md
index 59d1e97..82cb811 100644
--- a/doc/Advanced-Tutorial.md
+++ b/doc/Advanced-Tutorial.md
@@ -172,24 +172,7 @@ Starting with version 3.1.4, PartitionFinder supports the marginal Akaike Inform
If you use PartitionFinder-mAIC in a publication, please cite:
> TBD
-PartitionFinder implementation includes four merging algorithms, which differ in how thoroughly they search the space of partitioning schemes and how many pairs they merge per iteration:
-
-| Algorithm | IQ-TREE Command | Description | Reference |
-| ------------------------------------ | --------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------- | ----------------------- |
-| Greedy | `greedy` | At each iteration, evaluates all candidate subset pairs and merges the single pair that gives the best information-criterion score. | [Lanfear et al., 2012] |
-| Relaxed hierarchical clustering | `rcluster` | At each iteration, evaluates only the top k% most similar candidate subset pairs and merges the best-scoring one. | [Lanfear et al., 2014] |
-| Fast hierarchical relaxed clustering | `rclusterf` | At each iteration, evaluates only the top k% most similar candidate subset pairs and merges multiple compatible pairs at once. (IQ-TREE default). | [Lanfear et al., 2014] |
-| k-means | `kmeans` | Estimates an evolutionary rate for each site, then iteratively clusters individual sites by rate using k-means, without relying on predefined data blocks. | [Frandsen et al., 2015] |
-
-The following options control the details of the merging procedure:
-
-| Option | Description |
-| ---------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
-| `-merit` | Specify either `AIC`, `AICc`, `BIC` or `mAIC` for the partition merging criterion. *DEFAULT: `BIC`* |
-| `--merge` | Specify either `rclusterf`, `rcluster`, `greedy` or `kmeans` algorithm. *DEFAULT: `rclusterf`* |
-| `--rclusterf` | Specify the percentage of most similar candidate partition pairs retained for merging at each iteration (this automatically selects the fast relaxed clustering algorithm). *DEFAULT: 10* |
-| `--rcluster` | Specify the percentage of most similar candidate partition pairs retained for merging at each iteration (this automatically selects the relaxed clustering algorithm). *DEFAULT: 10* |
-| `--rcluster-max` | Specify the maximum number of most similar candidate partition pairs retained for merging at each iteration (only for the fast relaxed clustering and relaxed clustering algorithms). *DEFAULT: 10 × number of partitions in the full partition model* |
+The PartitionFinder implementation includes several algorithms. The options for controlling the details of partitioning scheme selection are documented in the [Automatic model selection](https://iqtree.github.io/doc/Command-Reference#automatic-model-selection) section of the Command Reference.
Ultrafast bootstrapping with partition model
--------------------------------------------
diff --git a/doc/Command-Reference.md b/doc/Command-Reference.md
index aa59d83..d22996f 100644
--- a/doc/Command-Reference.md
+++ b/doc/Command-Reference.md
@@ -204,8 +204,7 @@ Automatic model selection
-------------------------
-The default model (e.g., `HKY+F` for DNA, `LG` for protein data) may not fit well to the data. Therefore, IQ-TREE
-allows to automatically determine the best-fit model via a series of `-m TEST...` option:
+The default model (e.g., `HKY+F` for DNA, `LG` for protein data) may not fit well to the data. Therefore, IQ-TREE allows to automatically determine the best-fit model via a series of `-m TEST...` option:
|Option| Usage and meaning |
|----------------------|------------------------------------------------------------------------------|
@@ -231,33 +230,34 @@ IQ-TREE version 1.6 or later allows to additionally test [Lie Markov DNA models]
When [a partition file is specified](#partition-model-options) then you can append `MERGE` keyword into `-m` option to find the best-fit partitioning scheme like PartitionFinder ([Lanfear et al., 2012]). More specifically,
-|Option| Usage and meaning |
-|----------------------|------------------------------------------------------------------------------|
-| `-m TESTMERGEONLY` | Select best-fit partitioning scheme by possibly merging partitions to reduce over-parameterization and increase model fit. It implements the greedy algorithm of PartitionFinder. |
-| `-m TESTMERGE` | Like `-m TESTMERGEONLY` but immediately followed by tree reconstruction using the best partitioning scheme found. |
-| `-m TESTNEWMERGEONLY` or `-m MF+MERGE` | Like `-m TESTMERGEONLY` but additionally includes FreeRate model. |
-| `-m TESTNEWMERGE` or `-m MFP+MERGE` | Like `-m MF+MERGE` but immediately followed by tree reconstruction using the best partitioning scheme found. |
-| `-rcluster` | Specify the percentage for the relaxed clustering algorithm ([Lanfear et al., 2014]) to speed up the computation instead of the default slow greedy algorithm. This is similar to `--rcluster-percent` option of PartitionFinder. For example, with `-rcluster 10` only the top 10% partition schemes are considered to save computations. |
-| `-rclusterf` | Similar to `-rcluster` but using the **fast** relaxed clustering algorithm ([Lanfear et al., 2017]) of PartitionFinder2. Introduced in version 1.6. |
-| `-rcluster-max` | Specify the absolute maximum number of partition pairs in the paritition merging phase. Default: the larger of 1000 and 10 times the number of partitions. This option is similar to `--rcluster-max` option of PartitionFinder2. |
+| Option | Usage and meaning |
+| -------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
+| `-m TESTMERGEONLY` | Select best-fit partitioning scheme by possibly merging partitions to reduce over-parameterization and increase model fit. It implements the greedy algorithm of PartitionFinder. |
+| `-m TESTMERGE` | Like `-m TESTMERGEONLY` but immediately followed by tree reconstruction using the best partitioning scheme found. |
+| `-m TESTNEWMERGEONLY` or `-m MF+MERGE` | Like `-m TESTMERGEONLY` but additionally includes FreeRate model. |
+| `-m TESTNEWMERGE` or `-m MFP+MERGE` | Like `-m MF+MERGE` but immediately followed by tree reconstruction using the best partitioning scheme found. |
+| `--merge` | Specify either `rcluster` (relaxed clustering algorithm: at each iteration, evaluates only the top k% partition schemes and merges best one [Lanfear et al., 2014]), `rclusterf` (**fast** relaxed clustering algorithm: similar to `rcluster` but merges multiple compatible schemes at once [Lanfear et al., 2017]) or `greedy` (greedy algorithm: at each iteration, evaluates all candidate schemes and merges the best one [Lanfear et al., 2012]) algorithm. *DEFAULT: `rclusterf`* |
+| `-rcluster` | Specify the percentage for the relaxed clustering algorithm. This is similar to `--rcluster-percent` option of PartitionFinder. For example, with `-rcluster 10` only the top 10% partition schemes are evaluated. This automatically selects the relaxed clustering algorithm. *DEFAULT: 10* |
+| `-rclusterf` | Similar to `-rcluster` but using the **fast** relaxed clustering algorithm of PartitionFinder2. This automatically selects the fast relaxed clustering algorithm. Introduced in version 1.6. *DEFAULT: 10* |
+| `-rcluster-max` | Specify the absolute maximum number of partition pairs in the paritition merging phase. This option is similar to `--rcluster-max` option of PartitionFinder2. *DEFAULT: 10 times the number of partitions* |
> **WARNING**: For versions <= 1.5.X, all commands with `-m ...MERGE...` will always perform an edge-unlinked partition scheme finding even if `-spp` option is used. Only in the next phase of tree reconstruction, then an edge-linked partition model is used. However, for versions 1.6.X onwards, the edge-linked partition finding is performed by `-spp` option.
Several parameters can be set to e.g. reduce computations:
-|Option| Usage and meaning |
-|-------------|------------------------------------------------------------------------------|
-| `-mset` | Specify the name of a program (`raxml`, `phyml` or `mrbayes`) to restrict to only those models supported by the specified program. Alternatively, one can specify a comma-separated list of base models. For example, `-mset WAG,LG,JTT` will restrict model selection to WAG, LG, and JTT instead of all 18 AA models to save computations. |
-| `-msub` | Specify either `nuclear`, `mitochondrial`, `chloroplast` or `viral` to restrict to those AA models designed for specified source. |
-| `-mfreq` | Specify a comma-separated list of frequency types for model selection. *DEFAULT: `-mfreq FU,F` for protein models (FU = AA frequencies given by the protein matrix, F = empirical AA frequencies from the data), `-mfreq ,F1x4,F3x4,F` for codon models* |
-| `-mrate` | Specify a comma-separated list of rate heterogeneity types for model selection. *DEFAULT: `-mrate E,I,G,I+G` for standard procedure, `-mrate E,I,G,I+G,R` for new selection procedure*. (E means Equal/homogeneous rate model). |
-| `-cmin` | Specify minimum number of categories for FreeRate model. *DEFAULT: 2* |
-| `-cmax` | Specify maximum number of categories for FreeRate model. It is recommended to increase if alignment is long enough. *DEFAULT: 10* |
-| `-merit` | Specify either `AIC`, `AICc` or `BIC` for the optimality criterion to apply for new procedure. *DEFAULT: all three criteria are considered* |
-| `-mtree` | Turn on full tree search for each model considered, to obtain more accurate result. Only recommended if enough computational resources are available. *DEFAULT: fixed starting tree* |
-| `-mredo` | Ignore model checkpoint file computed earlier. *DEFAULT: model checkpoint file (if exists) is loaded to reuse previous computations* |
-| `-madd` | Specify a comma-separated list of mixture models to additionally consider for model selection. For example, `-madd LG4M,LG4X` to additionally include these two [protein mixture models](Substitution-Models#protein-models). |
-| `-mdef` | Specify a [NEXUS model file](Complex-Models#nexus-model-file) to define new models. |
+| Option | Usage and meaning |
+| -------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `-mset` | Specify the name of a program (`raxml`, `phyml` or `mrbayes`) to restrict to only those models supported by the specified program. Alternatively, one can specify a comma-separated list of base models. For example, `-mset WAG,LG,JTT` will restrict model selection to WAG, LG, and JTT instead of all 18 AA models to save computations. |
+| `-msub` | Specify either `nuclear`, `mitochondrial`, `chloroplast` or `viral` to restrict to those AA models designed for specified source. |
+| `-mfreq` | Specify a comma-separated list of frequency types for model selection. *DEFAULT: `-mfreq FU,F` for protein models (FU = AA frequencies given by the protein matrix, F = empirical AA frequencies from the data), `-mfreq ,F1x4,F3x4,F` for codon models* |
+| `-mrate` | Specify a comma-separated list of rate heterogeneity types for model selection. *DEFAULT: `-mrate E,I,G,I+G` for standard procedure, `-mrate E,I,G,I+G,R` for new selection procedure*. (E means Equal/homogeneous rate model). |
+| `-cmin` | Specify minimum number of categories for FreeRate model. *DEFAULT: 2* |
+| `-cmax` | Specify maximum number of categories for FreeRate model. It is recommended to increase if alignment is long enough. *DEFAULT: 10* |
+| `-merit` | Specify either `AIC`, `AICc` or `BIC` for the optimality criterion to apply for model selection. Addtionally, `mAIC` option ([Susko et al., 2026]) is available as partition model selection criterion. *DEFAULT: `BIC`* |
+| `-mtree` | Turn on full tree search for each model considered, to obtain more accurate result. Only recommended if enough computational resources are available. *DEFAULT: fixed starting tree* |
+| `-mredo` | Ignore model checkpoint file computed earlier. *DEFAULT: model checkpoint file (if exists) is loaded to reuse previous computations* |
+| `-madd` | Specify a comma-separated list of mixture models to additionally consider for model selection. For example, `-madd LG4M,LG4X` to additionally include these two [protein mixture models](Substitution-Models#protein-models). |
+| `-mdef` | Specify a [NEXUS model file](Complex-Models#nexus-model-file) to define new models. |
>**NOTE**: Some of the above options require a comma-separated list, which should not contain any empty space!
@@ -271,7 +271,7 @@ Several parameters can be set to e.g. reduce computations:
iqtree -s prot.phy -m MF -mset WAG,LG,JTT
-* Find the best partitioning scheme for alignment `data.phy` and partition file `partition.nex` with a relaxed clustering at 10% to save time:
+* Find the best partitioning scheme for alignment `data.phy` and partition file `partition.nex` with a relaxed clustering at 10%:
iqtree -s data.phy -spp partition.nex -m TESTMERGEONLY -rcluster 10
@@ -803,6 +803,7 @@ The first few lines of the output file example.phy.sitelh (printed by `-wslr` op
[Lanfear et al., 2012]: https://doi.org/10.1093/molbev/mss020
[Lanfear et al., 2014]: https://doi.org/10.1186/1471-2148-14-82
[Lanfear et al., 2017]: https://doi.org/10.1093/molbev/msw260
+[Susko et al., 2026]: https://doi.org/10.1093/sysbio/syag013
[Lartillot and Philippe, 2004]: https://doi.org/10.1093/molbev/msh112
[Ly-Trong et al., 2024]: https://doi.org/10.1093/molbev/msae134
[Minh et al., 2013]: https://doi.org/10.1093/molbev/mst024
From 81793826307afa21a07ae73d8206fd2a3f4e6fb8 Mon Sep 17 00:00:00 2001
From: HuaiyanRen <928003585@qq.com>
Date: Fri, 4 Sep 2026 19:24:04 +1000
Subject: [PATCH 08/10] Update Command-Reference.md
---
doc/Command-Reference.md | 20 ++++++++++----------
1 file changed, 10 insertions(+), 10 deletions(-)
diff --git a/doc/Command-Reference.md b/doc/Command-Reference.md
index d22996f..7d91974 100644
--- a/doc/Command-Reference.md
+++ b/doc/Command-Reference.md
@@ -230,16 +230,16 @@ IQ-TREE version 1.6 or later allows to additionally test [Lie Markov DNA models]
When [a partition file is specified](#partition-model-options) then you can append `MERGE` keyword into `-m` option to find the best-fit partitioning scheme like PartitionFinder ([Lanfear et al., 2012]). More specifically,
-| Option | Usage and meaning |
-| -------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
-| `-m TESTMERGEONLY` | Select best-fit partitioning scheme by possibly merging partitions to reduce over-parameterization and increase model fit. It implements the greedy algorithm of PartitionFinder. |
-| `-m TESTMERGE` | Like `-m TESTMERGEONLY` but immediately followed by tree reconstruction using the best partitioning scheme found. |
-| `-m TESTNEWMERGEONLY` or `-m MF+MERGE` | Like `-m TESTMERGEONLY` but additionally includes FreeRate model. |
-| `-m TESTNEWMERGE` or `-m MFP+MERGE` | Like `-m MF+MERGE` but immediately followed by tree reconstruction using the best partitioning scheme found. |
-| `--merge` | Specify either `rcluster` (relaxed clustering algorithm: at each iteration, evaluates only the top k% partition schemes and merges best one [Lanfear et al., 2014]), `rclusterf` (**fast** relaxed clustering algorithm: similar to `rcluster` but merges multiple compatible schemes at once [Lanfear et al., 2017]) or `greedy` (greedy algorithm: at each iteration, evaluates all candidate schemes and merges the best one [Lanfear et al., 2012]) algorithm. *DEFAULT: `rclusterf`* |
-| `-rcluster` | Specify the percentage for the relaxed clustering algorithm. This is similar to `--rcluster-percent` option of PartitionFinder. For example, with `-rcluster 10` only the top 10% partition schemes are evaluated. This automatically selects the relaxed clustering algorithm. *DEFAULT: 10* |
-| `-rclusterf` | Similar to `-rcluster` but using the **fast** relaxed clustering algorithm of PartitionFinder2. This automatically selects the fast relaxed clustering algorithm. Introduced in version 1.6. *DEFAULT: 10* |
-| `-rcluster-max` | Specify the absolute maximum number of partition pairs in the paritition merging phase. This option is similar to `--rcluster-max` option of PartitionFinder2. *DEFAULT: 10 times the number of partitions* |
+| Option | Usage and meaning |
+| -------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `-m TESTMERGEONLY` | Select best-fit partitioning scheme by possibly merging partitions to reduce over-parameterization and increase model fit. It implements the greedy algorithm of PartitionFinder. |
+| `-m TESTMERGE` | Like `-m TESTMERGEONLY` but immediately followed by tree reconstruction using the best partitioning scheme found. |
+| `-m TESTNEWMERGEONLY` or `-m MF+MERGE` | Like `-m TESTMERGEONLY` but additionally includes FreeRate model. |
+| `-m TESTNEWMERGE` or `-m MFP+MERGE` | Like `-m MF+MERGE` but immediately followed by tree reconstruction using the best partitioning scheme found. |
+| `--merge` | Specify either `rcluster` (relaxed clustering algorithm: at each iteration, evaluates only the top k% partition schemes and merges best one; [Lanfear et al., 2014]), `rclusterf` (**fast** relaxed clustering algorithm: similar to `rcluster` but merges multiple compatible schemes at once; [Lanfear et al., 2017]) or `greedy` (greedy algorithm: at each iteration, evaluates all candidate schemes and merges the best one; [Lanfear et al., 2012]) algorithm. *DEFAULT: `rclusterf`* |
+| `-rcluster` | Specify the percentage for the relaxed clustering algorithm. This is similar to `--rcluster-percent` option of PartitionFinder. For example, with `-rcluster 10` only the top 10% partition schemes are evaluated. This automatically selects the relaxed clustering algorithm. *DEFAULT: 10* |
+| `-rclusterf` | Similar to `-rcluster` but using the **fast** relaxed clustering algorithm of PartitionFinder2. This automatically selects the fast relaxed clustering algorithm. Introduced in version 1.6. *DEFAULT: 10* |
+| `-rcluster-max` | Specify the absolute maximum number of partition pairs in the paritition merging phase. This option is similar to `--rcluster-max` option of PartitionFinder2. *DEFAULT: 10 times the number of partitions* |
> **WARNING**: For versions <= 1.5.X, all commands with `-m ...MERGE...` will always perform an edge-unlinked partition scheme finding even if `-spp` option is used. Only in the next phase of tree reconstruction, then an edge-linked partition model is used. However, for versions 1.6.X onwards, the edge-linked partition finding is performed by `-spp` option.
From 312af4fe15ded9fa33315ec31dce83bcff7c3263 Mon Sep 17 00:00:00 2001
From: HuaiyanRen <928003585@qq.com>
Date: Thu, 10 Sep 2026 10:15:28 +1000
Subject: [PATCH 09/10] Update Advanced-Tutorial.md
---
doc/Advanced-Tutorial.md | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/doc/Advanced-Tutorial.md b/doc/Advanced-Tutorial.md
index 82cb811..e30780c 100644
--- a/doc/Advanced-Tutorial.md
+++ b/doc/Advanced-Tutorial.md
@@ -170,7 +170,7 @@ Starting with version 3.1.4, PartitionFinder supports the marginal Akaike Inform
iqtree -s example.phy -p example.nex -m MFP+MERGE -merit mAIC
If you use PartitionFinder-mAIC in a publication, please cite:
-> TBD
+> __H. Ren, T.K.F. Wong, C. Jiang, E. Susko, R. Lanfear, B.Q. Minh__ (2026) PartitionFinder-mAIC: Phylogenetic Partitioning using Marginal Akaike Information Criterion _bioRxiv_.
The PartitionFinder implementation includes several algorithms. The options for controlling the details of partitioning scheme selection are documented in the [Automatic model selection](https://iqtree.github.io/doc/Command-Reference#automatic-model-selection) section of the Command Reference.
From f27efeae350d263404070123d123e8982d15a258 Mon Sep 17 00:00:00 2001
From: HuaiyanRen <928003585@qq.com>
Date: Thu, 10 Sep 2026 10:19:35 +1000
Subject: [PATCH 10/10] Update Advanced-Tutorial.md
---
doc/Advanced-Tutorial.md | 1 -
1 file changed, 1 deletion(-)
diff --git a/doc/Advanced-Tutorial.md b/doc/Advanced-Tutorial.md
index e30780c..a0deb43 100644
--- a/doc/Advanced-Tutorial.md
+++ b/doc/Advanced-Tutorial.md
@@ -585,7 +585,6 @@ See [Command Reference](Command-Reference) for a complete list of all options av
[Lanfear et al., 2012]: https://doi.org/10.1093/molbev/mss020
[Susko et al., 2026]: https://doi.org/10.1093/sysbio/syag013
[Lanfear et al., 2014]: https://doi.org/10.1186/1471-2148-14-82
-[Frandsen et al., 2015]: https://doi.org/10.1186/s12862-015-0283-7
[Lopez et al., 2002]: http://mbe.oxfordjournals.org/content/19/1/1.full
[Nei et al., 2001]: https://doi.org/10.1073/pnas.051611498
[Seo et al., 2005]: https://doi.org/10.1073/pnas.0408313102