dna_data_structures.py

import collections
from typing import Dict, Union

class InvalidNucleotideError(Exception):
    """Custom errors for invalid nucleotides"""
    pass

class DNA:
    """"
    Represents a DNA sequence and provides basic opeations and statisitcs
    The sequence is stored as a string after validation
     """

    valid_nucleotides = ['A', 'T', 'C', 'G']

    def __init__(self, sequence):
        """"
        Initialise the DNA sequence
        Args:
        sequence (str): The DNA sequence
        Raises:
            InvalidNucleotideError if the sequence does not contain a valid nucleotide
        """
        #converting sequence to uppercase for validation and storage
        self.sequence = sequence.upper()
        self._validate_sequence()
        #pre-calculating counts for efficiency (using dict as required)
        self._nucleotide_counts= self._count_nucleotides()

    def _validate_sequence(self):
        """"
        Validates the DNA sequence, ensures only A,T,G,C nucleotides are allowed
        Handles invalid nucleotides with appropriate error message
        """
        #Implementing validation:
        invalid_nucleotides = set(self.sequence)- set(self.valid_nucleotides)

        if invalid_nucleotides:
            error_message= (f"Invalid nucleotides found in DNA sequence:"
                            f"{', '.join(sorted(invalid_nucleotides))}"
                            f"Only A,T,G,C nucleotides are allowed")
            raise InvalidNucleotideError(error_message)

    def _count_nucleotides(self) -> Dict[str, int]:
        """"
        Counts the nucleotides in the DNA sequence using collections.Counter
        Uses dictionary-like structure to store nucleotide counts
        """
        return collections.Counter(self.sequence)

    def get_length (self) -> int:
        """"
        Returns the length of the DNA sequence
        """
        return len(self.sequence)

    def get_nucleotide_count(self) -> Dict[str, int]:
        """"
        Returns count of each nucleotide in the DNA sequence
        """
        return self._nucleotide_counts

    def calculate_GC_content(self) ->float:
        """"
        Calculates and returns the percentage of  Guanine and Cytosine (GC) content
        """
        total_length = self.get_length()
        if total_length == 0:
            return 0.0
        G_count= self._nucleotide_counts.get('G', 0)
        C_count= self._nucleotide_counts.get('C', 0)

        GC_content=((G_count + C_count) / total_length) * 100
        return round(GC_content,2)

    def calculate_AT_content(self) -> float:
        """"
        Calculates and returns the percentage of Adenine and Thymine (AT) content
        """
        return round(100.0- self.calculate_GC_content(),2)

    def __str__(self):
        """"
        String representation of the DNA sequence
        """
        return f'DNA sequence: {self.sequence}'