PartitionFinder官网
http://www.robertlanfear.com/partitionfinder/

图片.png

1.下载：

1）下载后上传服务器。

图片.png

2) 直接下载到服务器

可以使用git clone 和wget命令下载

git clone https://github.com/brettc/partitionfinder.git
wget https://codeload.github.com/brettc/partitionfinder/tar.gz/refs/tags/v2.1.1

2.环境准备

图片.png

partitionfinder使用前准备Python2环境以及依赖包。python3是不行的。
现在一般都是Python3，所以我单独创建一个环境。

#创建Python2.7的环境
conda create -n partitionfinder python=2.7
#激活该环境
source activate partitionfinder
#安装依赖包
conda install numpy pandas pyparsing scipy 
pip install -U scikit-learn
pip install tables
#参考//www.greatytc.com/p/855bda1fb2c3

3.使用测试

进入环境后，看看是否能输出help。

source activate partitionfinder
python PartitionFinder.py -h

(partitionfinder) animal1@animalia:/apps/partitionfinder-2.1.1$ python PartitionFinder.py -h
INFO     | 2021-07-23 16:45:25,998 | Note: NumExpr detected 48 cores but "NUMEXPR_MAX_THREADS" not s
INFO     | 2021-07-23 16:45:25,998 | NumExpr defaulting to 8 threads.
Usage: python PartitionFinder.py [options] <foldername>

    PartitionFinder and PartitionFinderProtein are designed to discover optimal
    partitioning schemes for nucleotide and amino acid sequence alignments.
    They are also useful for finding the best model of sequence evolution for datasets.

    The Input: <foldername>: the full path to a folder containing:
        - A configuration file (partition_finder.cfg)
        - A nucleotide/aa alignment in Phylip format
    Take a look at the included 'example' folder for more details.

    The Output: A file in the same directory as the .cfg file, named
    'analysis' This file contains information on the best
    partitioning scheme, and the best model for each partiiton

    Usage Examples:
        >python PartitionFinder.py example
        Analyse what is in the 'example' sub-folder in the current folder.

        >python PartitionFinder.py -v example
        Analyse what is in the 'example' sub-folder in the current folder, but
        show all the debug output

        >python PartitionFinder.py -c ~/data/frogs
        Check the configuration files in the folder data/frogs in the current
        user's home folder.

        >python PartitionFinder.py --force-restart ~/data/frogs
        Deletes any data produced by the previous runs (which is in
        ~/data/frogs/output) and starts afresh


Options:
  -h, --help            show this help message and exit
  -v, --verbose         show debug logging information (equivalent to --debug-
                        out=all)
  -c, --check-only      just check the configuration files, don't do any
                        processing
  -f, --force-restart   delete all previous output and start afresh (!)
  -p N, --processes=N   Number of concurrent processes to use. Use -1 to match
                        the number of cpus on the machine. The default is to
                        use -1.
  --show-python-exceptions
                        If errors occur, print the python exceptions
  --save-phylofiles     save all of the phyml or raxml output. This can take a
                        lot of space(!)
  --dump-results        Dump all results to a binary file. This is only of use
                        for testing purposes.
  --compare-results     Compare the results to previously dumped binary
                        results. This is only of use for testing purposes.
  -q, --quick           Avoid anything slow (like writing schemes at each
                        step),useful for very large datasets.
  -r, --raxml           Use RAxML (rather than PhyML) to do the analysis. See
                        the manual
  -n, --no-ml-tree      Estimate a starting tree with NJ (PhyML) or MP (RaxML)
                        instead of the default which is to estimate a starting
                        tree with ML  using in RAxML. Not recommended.
  --cmdline-extras=N    Add additional commands to the phyml or raxml
                        commandlines that PF uses.This can be useful e.g. if
                        you want to change the accuracy of lnL calculations
                        ('-e' option in raxml), or use multi-threaded versions
                        of raxml that require you to specify the number of
                        threads you will let raxml use ('-T' option in raxml.
                        E.g. you might specify this: --cmndline_extras ' -e
                        2.0 -T 10 ' N.B. MAKE SURE YOU PUT YOUR EXTRAS IN
                        QUOTES, and only use this command if you really know
                        what you're doing and are very familiar with raxml and
                        PartitionFinder
  --weights=N           Mainly for algorithm development. Only use it if you
                        know what you're doing.A list of weights to use in the
                        clustering algorithms. This list allows you to assign
                        different weights to: the overall rate for a subset,
                        the base/amino acid frequencies, model parameters, and
                        alpha value. This will affect how subsets are
                        clustered together. For instance: --cluster_weights
                        '1, 2, 5, 1', would weight the base freqeuncies 2x
                        more than the overall rate, the model parameters 5x
                        more, and the alpha parameter the same as the model
                        rate
  --kmeans=type         This defines which sitewise values to use: entropy or
                        tiger  --kmeans entropy: use entropies for sitewise
                        values --kmeans tiger: use TIGER rates for sitewise
                        values (only valid for Morphology)
  --rcluster-percent=N  This defines the proportion of possible schemes that
                        the relaxed clustering algorithm will consider before
                        it stops looking. The default is 10%. e.g. --rcluster-
                        percent 10.0
  --rcluster-max=N      This defines the number of possible schemes that the
                        relaxed clustering algorithm will consider before it
                        stops looking. The default is to look at the larger
                        value out of 1000, and 10 times the number of data
                        blocks you have. e.g. --rcluster-max 1000
  --min-subset-size=N   This defines the minimum subset size that the kmeans
                        and rcluster algorithm will accept. Subsets smaller
                        than this  will be merged at with other subsets at the
                        end of the algorithm (for kmeans) or at the start of
                        the algorithm (for rcluster). See manual for details.
                        The default value for kmeans is 100. The default value
                        for rcluster is to ignore this option. e.g. --min-
                        subset-size 100
  --debug-output=REGION,REGION,...
                        (advanced option) Provide a list of debug regions to
                        output extra information about what the program is
                        doing. Possible regions are 'all' or any of {subset,su
                        bset_ops,raxml,parser,model_util,results,entropy,numex
                        pr,alignment,concurrent.futures,threadpool,numexpr.uti
                        ls,progress,main,config,reporter,kmeans,util,concurren
                        t,morph_tige,analysis_m,neighbour,scheme,submodels,dat
                        abase,analysis,phyml,raxml_mode,model_load,phyml_mode,
                        sklearn}.
  --all-states          In the kmeans and rcluster algorithms, this stipulates
                        that PartitionFinder should not produce subsets that
                        do not have all possible states present. E.g. for DNA
                        sequence data, all subsets in the final scheme must
                        have A, C, T,  and G nucleotides present. This can
                        occasionally be useful for downstream  analyses,
                        particularly concerning amino acid datasets.
  --profile             Output profiling information after running (this will
                        slow everything down!)

3.使用方法

1）准备序列矩阵文件以及配置文件

准备一个文件夹下包含phy文件和cfg。
phy是序列矩阵信息，cfg是配置文件

图片.png

## ALIGNMENT FILE 序列矩阵文件##
alignment = Acan.phy;

## BRANCHLENGTHS: linked | unlinked （一般）##
branchlengths = unlinked;

## MODELS OF EVOLUTION: all | allx | mrbayes | beast | gamma | gammai | <list> ##
models = mrbayes;

# MODEL SELECCTION: AIC | AICc | BIC #
model_selection = bic;

## DATA BLOCKS: see manual for how to define （分区情况）##
[data_blocks]
atp6 = 1-107;
cox1 = 108-566;
cox2 = 567-731;
cox3 = 732-870;
cytb = 871-1182;
nad1 = 1183-1382;
nad2 = 1383-1523;
nad3 = 1524-1574;
nad4L = 1575-1599;
nad4 = 1600-1818;
nad5 = 1819-2055;
nad6 = 2056-2087;

## SCHEMES, search: all | user | greedy | rcluster | rclusterf | kmeans ##
[schemes]
search = greedy;

2）运行

序列矩阵文件以及配置文件置于一文件夹下

(partitionfinder) animal1@animalia:~/Documents/20210723_MB/PartitionFinder$ l
Acan.phy  partition_finder.cfg
(partitionfinder) animal1@animalia:~/Documents/20210723_MB/PartitionFinder$ cd ../
(partitionfinder) animal1@animalia:~/Documents/20210723_MB$ l
PartitionFinder/

运行方式：
python <$PartitionFinder文件路径/PartitionFinder.py> <序列矩阵文件以及配置文件的文件夹>
注意！
氨基酸序列分析使用PartitionFinderProtein.py
核苷酸序列分析使用PartitionFinder.py
我这里使用氨基酸序列进行分析。

python /apps/partitionfinder-2.1.1/PartitionFinderProtein.py full_path/PartitionFinder

主要的可用运行结果在analysis/schemes/start_scheme.txt 文件中 MrBayes 中各分区的适用模型。