Python wrapper for the bioinformatics tool mothur
Copyright © 2018, Richard Campen. All rights reserved.
See LICENSE.txt for full license conditions.
A python wrapper for the command line version of the bioinformatics tool mothur.
See the change log at the end of this ReadMe for a full list of changes.
Mothur-py was inspired by the ipython-mothurmagic module, but with an intention to provide a more general python wrapper that would work outside of the IPython/Jupyter notebook environment, as well as provide support for mothur’s current keyword functionality.
Note: This module has only been tested with mothur v1.39.5 and python 3. It should in theory work with older versions of mothur, but the older the version the less likely as this module relies upon some of the more recent mothur commands/output to function properly.
To install the latest release version you can just pip install mothur-py. To install the most up to date code you should download/clone this repository and create a binary distribution using python setup.py bdist_wheel that will create a .whl file in the dist folder. You can then install mothur-py with pip from the .whl file using pip install <wheel_file_name>. The advantage of this method over just running python setup.py install is that you can easily remove or update the package via pip.
NOTE: mothur-py expects mothur to be installed in the users PATH environment variable. If this is not the case you will need to tell it where to find the mothur executable. See the configuration section of the README for details.
Use of this module revolves around the Mothur class that catches method calls and passes them off to mothur to be run as commands. An instance of the Mothur class needs to be created before running any commands:
# create instance of Mothur class from mothur_py import Mothur m = Mothur()
Commands in mothur can then be executed as methods of the Mothur class instance using the same names you would use within the command line version of mothur:
# run the mothur help command m.help()
Command parameters can either be passed as python native types (i.e. strings, integers, floats, booleans, lists) or as strings that match the format that mothur would expect:
# running make contigs using str input for file parameter, and int for processor paramenter m.make.contigs(file='basic_usage.files', processors=2) # running summary.single, passing calculators as mothur formatted list m.summary.single(shared='basic_usage.shared', calc='nseqs-sobs-coverage-shannon-simpson') # running summary.single, passing calculators as python list also works m.summary.single(shared='basic_usage.shared', calc=['nseqs', 'sobs', 'coverage', 'shannon', 'simpson'])
The Mothur object saves a record of the current directories and files and the output files from mothur after executing each command. These are stored as dictionary attributes of the Mothur object:
# run a command m.summary.seqs(fasta='basic_usage.fasta') # display current mothur files print(m.current_files) # read in the output file from summary.seqs() with open(m.output_files['summary'], 'r') as in_handle: in_handle.read()
NOTE: Due to the possibility of multiple output files with the same extension the output files are saved as lists within the attribute dictionaries with the file extension as the key. This issue does not occur for current files and dirs so they are stored as the actual values, not as lists of the values, with the key being the type of file according to mothur (usually the same as the file extension).
NOTE: Each successive execution of a mothur command will update the current files and dirs, but will completely overwrite the saved output files. This is so that you have access to the current files generated more than one command ago, but do not get access to output from more than one command ago, which would be confusing.
There is also implementation of the current keyword used in the command line version of mothur:
# run the mothur summary.seqs command using the 'current' option # NOTE: current is being passed as a string m.summary.seqs(fasta='current') # like the command line version, you don't even need to specify # the 'current' keyword for some commands m.summary.seqs()
Behind the scenes, the current keyword is enabled by appending the users command with the get.current() command to list the current directories and files being used by mothur, parsing of the output to extract this information, and prepending future commands with set.dir() and set.current() to tell mothur what these should be. This is necessary as each call to mothur is executed as a separate mothur session and therefore mothur can not store this information itself.
The Mothur class stores configuration options for how mothur is executed. These options include mothur_path to tell mothur-py where to find the mothur executable, verbosity to control how much output there is, mothur_seed to control the seed used by mothur for random number generation, logfile_name to set the name of the mothur logfile, suppress_logfile which suppresses the creation of the mothur logfile, and line_limit which sets the limit on how many lines of stdout will be printed.
The default for mothur_path is mothur which will only work if mothur is in your PATH environment variable. If it is not then you will need to specify where to find the mothur executable, including the name of the executable itself:
# configure mothur with executable in current directory on Windows m = Mothur(mothur_path='mothur.exe') # configure mothur with executable in current directory on Linux m = Mothur(mothur_path='./mothur') # configure mothur with executable in alternate directory on Windows m = Mothur(mothur_path='\\path\\to\\mothur.exe')
Failure to correctly configure the mothur_path will usually result in a PermissionError:
m = Mothur(mothur_path='/not/a/real/path/to/mothur') Traceback (most recent call last): File "<stdin>", line 1, in <module> File ".../mothur_py/core.py", line 199, in __call__ p = Popen([self.root_object.mothur_path, '#%s' % commands_str], stdout=PIPE, stderr=STDOUT) File "/usr/lib/python3.5/subprocess.py", line 947, in __init__ restore_signals, start_new_session) File "/usr/lib/python3.5/subprocess.py", line 1551, in _execute_child raise child_exception_type(errno_num, err_msg) PermissionError: [Errno 13] Permission denied
When verbosity is set to 0 (default) there is no output printed, 1 prints the normal output as would be seen with command line execution (minus the header that contains the mothur version and runtime information), and 2 displays all output including the commands being executed behind the scenes to enable the current keyword to work. The default option is 0, with 1 being useful when you want to see the standard mothur output, and 2 being useful for debugging purposes.
If mothur_seed is set to a valid integer then this number will be passed to mothur to be used for random number generation. This is implemented by adding the seed=<your seed here> named parameter to each mothur command. By default no seed is set (mothur_seed=None). Not all commands will accept having a seed set. For these commands you may need to set the mothur_seed parameter to None for the execution of that command, e.g.:
m = Mothur(mothur_seed=12345) # summary.seqs() allows setting the seed so this will run fine m.summary.seqs(fasta='current') # help() does not accept having the seed set so need to alter that value temporarily, otherwise an error will occur seed = m.mothur_seed m.mothur_seed = None m.help() m.mothur_seed = seed
The logfile_name option allows the user to specify the name of the mothur generated logfile. The logfile will store the output from all mothur commands executed for the Mothur object it is configured for. When set to None (default) a random logfile name is generated for the mothur object in the format mothur.py.<random_5_digit_number>.logfile.
Note: When copying mothur objects it is important to then specify different logfiles for them otherwise they may attempt to use the same logfile. Additionally, if suppress_logfile is true, the logfile will be suppressed even if it has been given a name by the user.
The supress_logfile option is useful when you don’t want the log files, such as when running in an Jupyter (nee IPython) notebook with verbosity=1, in which case you already have a record of mothur’s output and the mothur logfiles are superfluous. Default setting is False.
Note: Currently, due to the way that mothur creates the logfiles, a logfile will always be created BUT it will be cleaned up upon successful execution if suppress_logfile=True. However, if mothur fails to successfully execute, i.e. execution hangs or is interrupted, the logfile will not be cleaned up. For relevant discussion of this behaviour in mothur see here and here.
The line_limit option is useful when the full stdout is not wanted, or when printing it will be problematic, such as when stdout is excessive and causes memory issues in the Jupyter (nee IPython) notebook environment. Setting line_limit to -1 (default) imposes no line limit, while any positive integer (or zero) imposes a line limit that causes stdout to no longer be printed once the limit is reaches. If the line limit is reached then a warning is printed to notify the user.
Note: Only lines related to the user specified command count towards the line limit, therefore commands running in the background (viewable with verbosity == 2) do not count towards this limit. Additionally, when verbosity == 2 the additional commands that enable the current keyword functionality that are executed after the users command are still displayed, even if the line limit has been reached. Setting a line limit does not change what is printed to the logfile.
You can also instantiate the Mothur object with your desired configuration options.
m = Mothur(verbosity=1, mothur_seed=543210, suppress_logfile=True, line_limit=1000)
The current files and current directories for use in mothur are stored in dictionary attributes of the Mothur instance, current_files and current_dirs respectively. These values can be passed to mothur commands, e.g:
# passing current fasta file to summary.seqs() m.summary.seqs(fasta=m.current_files['fasta'])
The current keyword is actually just a shortcut for this functionality so it will always be easier to just pass 'current'. However, this demonstrates that the parameters of the mothur commands can accept any variable as long as it will resolve to something that mothur accepts. In the above example, the dictionary value resolves to a string that is the path to the .fasta file. As a better example of passing python variables as mothur command parameters, you could perform classification of sequences at multiple defined cutoffs as follows:
from copy import deepcopy # init results container mothur_objs = dict() # iterate over list off possible cutoff values for cutoff in [70, 80, 90]: # make a copy of the original mothur object so we do not make unwanted modifications to the original m_copy = deepcopy(m) # save outputs to different folders, but keep input the same output_dir = 'cutoff_%s' % cutoff m_copy.set.dir(output=output_dir, input='.') m_copy.classify.seqs(fasta='current', count='current', reference='reference.fasta', taxonomy='referenece.tax', cutoff=cutoff) # save to results container mothur_objs[cutoff] = m_copy
This may be a convoluted example, but it demonstrates the functionality well. One note of caution with this approach is that depending on the mothur command and the parameter you are changing, you may be overwriting your output files as you go. This is the reason for saving each output to a different folder in the above example. We also create multiple copies of the original mothur object and use those for the command instead so we can continue to use the current keyword downstream and have it refer to the correct files:
# using the results container from above for m_ in mothur_objs.values(): # run remove.lineage() on each mothur object created previously to remove unwanted taxa # because we saved each classify.seqs command to a different mothur object we can safely use the 'current' # keyword here and know that it refers to the correct file m_.remove.lineage(fasta='current', count='current', taxonomy='current, taxon='unknown')
You can also instantiate a Mothur instance with predefined current file and directory dictionaries:
m = Mothur(current_files=my_predefined_files_dict, current_dirs=my_predefined_files_dict)
This can be convenient for saving and loading the state of a mothur object to/from file as such:
import json # save state of mothur object, m, to json file with open('mothur_object.json', 'w') as out_handle: json.dump(vars(m), out_handle) # can reload mothur object from the json file with open('mothur_object.json', 'r') as in_handle: m = Mothur(**json.load(in_handle))
New features: * Added a line limit configuration option to limit the amount of stdout printed to screen. Helps with potential memory issues when outputting in Jupyter (nee IPython) notebook
Big fixes: * Current files/dirs and output are now only set after successful execution to prevent them being changed when mothur errors
Changes: * Added lines of text specifying transition from user input functions from background functions enabling the current keyword functionality when verbosity == 2 * For verbosity == 1, now does not print redundant warning/error text at the end of stdout * Split utility functions out into their own file (mothur_py.utils) * Refined __str__ and __repr__ for Mothur and MothurCommand
New features: * Allow setting the name of the mothur logfile via configuration of the Mothur object
Bug fixes: * Changed strings to match to detect errors and warnings from mothur in stdout * Now check both top level and output directories when removing logfile
Changes: * The output_files attribute now stores the full file path for each output file, rather than just the file name as was done previously.
New features: * Added configuration option for path to the mothur executable, thereby adding support for any mothur executable location (previously mothur had to be in users PATH environment variable).
New features: * Enabled passing python native types as command parameters, which are then converted to mothur compatible types as needed * Added parsing and saving of the output files generated by the last run mothur command * Improved documentation and examples
Bug fixes: * Fixed current files not being saved correctly to the mothur obejct after execution
Changes: * No longer return the mothur object. If it is desired to store the altered object as a copy then the deepcopy function in the copy package should be used
Bug fixes: * Fixed verbosity level affecting the parsing of current files/dirs from stdout * Only update the current files/dirs for the mothur object once execution of mothur ends successfully * Fixed calls to unimplemented dunder methods (i.e. deepcopy) being parsed as mothur commands
New features: * Can set the seed that mothur uses for its random number generation * Added more unittests
Bug fixes: * Fixed bug where python failed to raise an error if mothur did
New features: * Use of an invalid command now raises an error in python, halting execution. Previously this would fail silently.
Changes: * Renamed project from Rhea to mothur_py to avoid confusion with the R package for 16S amplicon analysis. Mothur class now needs to be imported from mothur_py instead.