#!/usr/bin/env python3
"""Module containing the Puckering class and the command line interface."""
import argparse
from typing import Optional
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
from biobb_common.configuration import settings
from biobb_common.generic.biobb_object import BiobbObject
from biobb_common.tools.file_utils import launchlogger
from biobb_dna.utils.common import _from_string_to_list
from biobb_dna.utils.loader import read_series
from biobb_dna.utils.transform import inverse_complement
[docs]
class Puckering(BiobbObject):
"""
| biobb_dna Puckering
| Calculate Puckering from phase parameters.
| Calculate North/East/West/South distribution of sugar puckering backbone torsions.
Args:
input_phaseC_path (str): Path to .ser file for helical parameter 'phaseC'. File type: input. `Sample file <https://raw.githubusercontent.com/bioexcel/biobb_dna/master/biobb_dna/test/data/backbone/canal_output_phaseC.ser>`_. Accepted formats: ser (edam:format_2330).
input_phaseW_path (str): Path to .ser file for helical parameter 'phaseW'. File type: input. `Sample file <https://raw.githubusercontent.com/bioexcel/biobb_dna/master/biobb_dna/test/data/backbone/canal_output_phaseW.ser>`_. Accepted formats: ser (edam:format_2330).
output_csv_path (str): Path to .csv file where output is saved. File type: output. `Sample file <https://raw.githubusercontent.com/bioexcel/biobb_dna/master/biobb_dna/test/reference/backbone/puckering_ref.csv>`_. Accepted formats: csv (edam:format_3752).
output_jpg_path (str): Path to .jpg file where output is saved. File type: output. `Sample file <https://raw.githubusercontent.com/bioexcel/biobb_dna/master/biobb_dna/test/reference/backbone/puckering_ref.jpg>`_. Accepted formats: jpg (edam:format_3579).
properties (dict):
* **sequence** (*str*) - (None) Nucleic acid sequence corresponding to the input .ser file. Length of sequence is expected to be the same as the total number of columns in the .ser file, minus the index column (even if later on a subset of columns is selected with the *seqpos* option).
* **stride** (*int*) - (1000) granularity of the number of snapshots for plotting time series.
* **seqpos** (*list*) - (None) list of sequence positions (columns indices starting by 0) to analyze. If not specified it will analyse the complete sequence.
* **remove_tmp** (*bool*) - (True) [WF property] Remove temporal files.
* **restart** (*bool*) - (False) [WF property] Do not execute if output files exist.
* **sandbox_path** (*str*) - ("./") [WF property] Parent path to the sandbox directory.
Examples:
This is a use example of how to use the building block from Python::
from biobb_dna.backbone.puckering import puckering
prop = {
'sequence': 'GCAT',
}
puckering(
input_phaseC_path='/path/to/phaseC.ser',
input_phaseW_path='/path/to/phaseW.ser',
output_csv_path='/path/to/table/output.csv',
output_jpg_path='/path/to/table/output.jpg',
properties=prop)
Info:
* wrapped_software:
* name: In house
* license: Apache-2.0
* ontology:
* name: EDAM
* schema: http://edamontology.org/EDAM.owl
"""
def __init__(
self,
input_phaseC_path,
input_phaseW_path,
output_csv_path,
output_jpg_path,
properties=None,
**kwargs,
) -> None:
properties = properties or {}
# Call parent class constructor
super().__init__(properties)
self.locals_var_dict = locals().copy()
# Input/Output files
self.io_dict = {
"in": {
"input_phaseC_path": input_phaseC_path,
"input_phaseW_path": input_phaseW_path,
},
"out": {
"output_csv_path": output_csv_path,
"output_jpg_path": output_jpg_path,
},
}
self.properties = properties
self.sequence = properties.get("sequence")
self.stride = properties.get("stride", 1000)
self.seqpos = [
int(elem) for elem in _from_string_to_list(properties.get("seqpos", None))
]
# Check the properties
self.check_properties(properties)
self.check_arguments()
[docs]
@launchlogger
def launch(self) -> int:
"""Execute the :class:`Puckering <backbone.puckering.Puckering>` object."""
# Setup Biobb
if self.check_restart():
return 0
self.stage_files()
# check sequence
if self.sequence is None or len(self.sequence) < 2:
raise ValueError("sequence is null or too short!")
# check seqpos
if self.seqpos:
if (max(self.seqpos) > len(self.sequence) - 2) or (min(self.seqpos) < 1):
raise ValueError(
f"seqpos values must be between 1 and {len(self.sequence) - 2}"
)
if not (isinstance(self.seqpos, list) and len(self.seqpos) > 1):
raise ValueError("seqpos must be a list of at least two integers")
else:
self.seqpos = None # type: ignore
# read input files
phaseC = read_series(
self.stage_io_dict["in"]["input_phaseC_path"], usecols=self.seqpos
)
phaseW = read_series(
self.stage_io_dict["in"]["input_phaseW_path"], usecols=self.seqpos
)
# fix angle range so its not negative
phaseC = self.fix_angles(phaseC)
phaseW = self.fix_angles(phaseW)
# calculate difference between epsil and zeta parameters
xlabels = self.get_xlabels(self.sequence, inverse_complement(self.sequence))
Npop, Epop, Wpop, Spop = self.check_puckering(phaseC, phaseW)
# save plot
fig, axs = plt.subplots(1, 1, dpi=300, tight_layout=True)
axs.bar(range(len(xlabels)), Npop, label="North")
axs.bar(range(len(xlabels)), Epop, bottom=Npop, label="East")
axs.bar(range(len(xlabels)), Spop, bottom=Npop + Epop, label="South")
axs.bar(range(len(xlabels)), Wpop, bottom=Npop + Epop + Spop, label="West")
# empty bar to divide both sequences
axs.bar([len(phaseC.columns)], [100], color="white", label=None)
axs.legend()
axs.set_xticks(range(len(xlabels)))
axs.set_xticklabels(xlabels, rotation=90)
axs.set_xlabel("Nucleotide Sequence")
axs.set_ylabel("Puckering (%)")
axs.set_title("Nucleotide parameter: Puckering")
fig.savefig(self.stage_io_dict["out"]["output_jpg_path"], format="jpg")
# save table
populations = pd.DataFrame(
{
"Nucleotide": xlabels,
"North": Npop,
"East": Epop,
"West": Wpop,
"South": Spop,
}
)
populations.to_csv(self.stage_io_dict["out"]["output_csv_path"], index=False)
plt.close()
# Copy files to host
self.copy_to_host()
# Remove temporary file(s)
# self.tmp_files.extend([self.stage_io_dict.get("unique_dir", "")])
self.remove_tmp_files()
self.check_arguments(output_files_created=True, raise_exception=False)
return 0
[docs]
def get_xlabels(self, strand1, strand2):
# get list of tetramers, except first and last two bases
labelsW = list(strand1)
labelsW[0] = f"{labelsW[0]}5'"
labelsW[-1] = f"{labelsW[-1]}3'"
labelsW = [f"{i}-{j}" for i, j in zip(labelsW, range(1, len(labelsW) + 1))]
labelsC = list(strand2)[::-1]
labelsC[0] = f"{labelsC[0]}5'"
labelsC[-1] = f"{labelsC[-1]}3'"
labelsC = [f"{i}-{j}" for i, j in zip(labelsC, range(len(labelsC), 0, -1))]
if self.seqpos:
labelsC = [labelsC[i] for i in self.seqpos]
labelsW = [labelsW[i] for i in self.seqpos]
xlabels = labelsW + ["-"] + labelsC
return xlabels
[docs]
def check_puckering(self, phaseC, phaseW):
separator_df = pd.DataFrame({"-": np.nan}, index=range(1, len(phaseC)))
phase = pd.concat([phaseW, separator_df, phaseC[phaseC.columns[::-1]]], axis=1)
# phase.columns = columns
Npop = np.logical_or(phase > 315, phase < 45).mean() * 100
Epop = np.logical_and(phase > 45, phase < 135).mean() * 100
Wpop = np.logical_and(phase > 225, phase < 315).mean() * 100
Spop = np.logical_and(phase > 135, phase < 225).mean() * 100
return Npop, Epop, Wpop, Spop
[docs]
def fix_angles(self, dataset):
values = np.where(dataset < 0, dataset + 360, dataset)
# values = np.where(values > 360, values - 360, values)
dataset = pd.DataFrame(values)
return dataset
[docs]
def puckering(
input_phaseC_path: str,
input_phaseW_path: str,
output_csv_path: str,
output_jpg_path: str,
properties: Optional[dict] = None,
**kwargs,
) -> int:
"""Create :class:`Puckering <dna.backbone.puckering.Puckering>` class and
execute the: meth: `launch() <dna.backbone.puckering.Puckering.launch>` method."""
return Puckering(
input_phaseC_path=input_phaseC_path,
input_phaseW_path=input_phaseW_path,
output_csv_path=output_csv_path,
output_jpg_path=output_jpg_path,
properties=properties,
**kwargs,
).launch()
puckering.__doc__ = Puckering.__doc__
[docs]
def main():
"""Command line execution of this building block. Please check the command line documentation."""
parser = argparse.ArgumentParser(
description="Calculate North/East/West/South distribution of sugar puckering backbone torsions.",
formatter_class=lambda prog: argparse.RawTextHelpFormatter(prog, width=99999),
)
parser.add_argument("--config", required=False, help="Configuration file")
required_args = parser.add_argument_group("required arguments")
required_args.add_argument(
"--input_phaseC_path",
required=True,
help="Helical parameter <alphaC> input ser file path. Accepted formats: ser.",
)
required_args.add_argument(
"--input_phaseW_path",
required=True,
help="Helical parameter <alphaW> input ser file path. Accepted formats: ser.",
)
required_args.add_argument(
"--output_csv_path",
required=True,
help="Path to output csv file. Accepted formats: csv.",
)
required_args.add_argument(
"--output_jpg_path",
required=True,
help="Path to output jpg file. Accepted formats: jpg.",
)
args = parser.parse_args()
args.config = args.config or "{}"
properties = settings.ConfReader(config=args.config).get_prop_dic()
puckering(
input_phaseC_path=args.input_phaseC_path,
input_phaseW_path=args.input_phaseW_path,
output_csv_path=args.output_csv_path,
output_jpg_path=args.output_jpg_path,
properties=properties,
)
if __name__ == "__main__":
main()