ColiFormer / CodonTransformer /CodonEvaluation.py
saketh11's picture
Add local CodonTransformer modules for custom ColiFormer functionality
6e9b5dc
"""
File: CodonEvaluation.py
---------------------------
Includes functions to calculate various evaluation metrics along with helper
functions.
"""
from typing import Dict, List, Tuple, Optional
import pandas as pd
from CAI import CAI, relative_adaptiveness
from tqdm import tqdm
import math
import numpy as np
from collections import Counter
from itertools import chain
from statistics import mean
import sys
import os
from io import StringIO
def get_CSI_weights(sequences: List[str]) -> Dict[str, float]:
"""
Calculate the Codon Similarity Index (CSI) weights for a list of DNA sequences.
Args:
sequences (List[str]): List of DNA sequences.
Returns:
dict: The CSI weights.
"""
return relative_adaptiveness(sequences=sequences)
def get_CSI_value(dna: str, weights: Dict[str, float]) -> float:
"""
Calculate the Codon Similarity Index (CSI) for a DNA sequence.
Args:
dna (str): The DNA sequence.
weights (dict): The CSI weights from get_CSI_weights.
Returns:
float: The CSI value.
"""
return CAI(dna, weights)
def get_organism_to_CSI_weights(
dataset: pd.DataFrame, organisms: List[str]
) -> Dict[str, dict]:
"""
Calculate the Codon Similarity Index (CSI) weights for a list of organisms.
Args:
dataset (pd.DataFrame): Dataset containing organism and DNA sequence info.
organisms (List[str]): List of organism names.
Returns:
Dict[str, dict]: A dictionary mapping each organism to its CSI weights.
"""
organism2weights = {}
# Iterate through each organism to calculate its CSI weights
for organism in tqdm(organisms, desc="Calculating CSI Weights: ", unit="Organism"):
organism_data = dataset.loc[dataset["organism"] == organism]
sequences = organism_data["dna"].to_list()
weights = get_CSI_weights(sequences)
organism2weights[organism] = weights
return organism2weights
def get_GC_content(dna: str) -> float:
"""
Calculate the GC content of a DNA sequence.
Args:
dna (str): The DNA sequence.
Returns:
float: The GC content as a percentage.
"""
dna = dna.upper()
if not dna:
return 0.0
return (dna.count("G") + dna.count("C")) / len(dna) * 100
def get_cfd(
dna: str,
codon_frequencies: Dict[str, Tuple[List[str], List[float]]],
threshold: float = 0.3,
) -> float:
"""
Calculate the codon frequency distribution (CFD) metric for a DNA sequence.
Args:
dna (str): The DNA sequence.
codon_frequencies (Dict[str, Tuple[List[str], List[float]]]): Codon
frequency distribution per amino acid.
threshold (float): Frequency threshold for counting rare codons.
Returns:
float: The CFD metric as a percentage.
"""
# Get a dictionary mapping each codon to its normalized frequency
codon2frequency = {
codon: freq / max(frequencies)
for amino, (codons, frequencies) in codon_frequencies.items()
for codon, freq in zip(codons, frequencies)
}
cfd = 0
# Iterate through the DNA sequence in steps of 3 to process each codon
for i in range(0, len(dna), 3):
codon = dna[i : i + 3]
codon_frequency = codon2frequency[codon]
if codon_frequency < threshold:
cfd += 1
return cfd / (len(dna) / 3) * 100
def get_min_max_percentage(
dna: str,
codon_frequencies: Dict[str, Tuple[List[str], List[float]]],
window_size: int = 18,
) -> List[float]:
"""
Calculate the %MinMax metric for a DNA sequence.
Args:
dna (str): The DNA sequence.
codon_frequencies (Dict[str, Tuple[List[str], List[float]]]): Codon
frequency distribution per amino acid.
window_size (int): Size of the window to calculate %MinMax.
Returns:
List[float]: List of %MinMax values for the sequence.
Credit: https://github.com/chowington/minmax
"""
# Get a dictionary mapping each codon to its respective amino acid
codon2amino = {
codon: amino
for amino, (codons, frequencies) in codon_frequencies.items()
for codon in codons
}
min_max_values = []
codons = [dna[i : i + 3] for i in range(0, len(dna), 3)] # Split DNA into codons
# Iterate through the DNA sequence using the specified window size
for i in range(len(codons) - window_size + 1):
codon_window = codons[i : i + window_size] # Codons in the current window
Actual = 0.0 # Average of the actual codon frequencies
Max = 0.0 # Average of the min codon frequencies
Min = 0.0 # Average of the max codon frequencies
Avg = 0.0 # Average of the averages of all frequencies for each amino acid
# Sum the frequencies for codons in the current window
for codon in codon_window:
aminoacid = codon2amino[codon]
frequencies = codon_frequencies[aminoacid][1]
codon_index = codon_frequencies[aminoacid][0].index(codon)
codon_frequency = codon_frequencies[aminoacid][1][codon_index]
Actual += codon_frequency
Max += max(frequencies)
Min += min(frequencies)
Avg += sum(frequencies) / len(frequencies)
# Divide by the window size to get the averages
Actual = Actual / window_size
Max = Max / window_size
Min = Min / window_size
Avg = Avg / window_size
# Calculate %MinMax
percentMax = ((Actual - Avg) / (Max - Avg)) * 100
percentMin = ((Avg - Actual) / (Avg - Min)) * 100
# Append the appropriate %MinMax value
if percentMax >= 0:
min_max_values.append(percentMax)
else:
min_max_values.append(-percentMin)
# Populate the last floor(window_size / 2) entries of min_max_values with None
for i in range(int(window_size / 2)):
min_max_values.append(None)
return min_max_values
def get_sequence_complexity(dna: str) -> float:
"""
Calculate the sequence complexity score of a DNA sequence.
Args:
dna (str): The DNA sequence.
Returns:
float: The sequence complexity score.
"""
def sum_up_to(x):
"""Recursive function to calculate the sum of integers from 1 to x."""
if x <= 1:
return 1
else:
return x + sum_up_to(x - 1)
def f(x):
"""Returns 4 if x is greater than or equal to 4, else returns x."""
if x >= 4:
return 4
elif x < 4:
return x
unique_subseq_length = []
# Calculate unique subsequences lengths
for i in range(1, len(dna) + 1):
unique_subseq = set()
for j in range(len(dna) - (i - 1)):
unique_subseq.add(dna[j : (j + i)])
unique_subseq_length.append(len(unique_subseq))
# Calculate complexity score
complexity_score = (
sum(unique_subseq_length) / (sum_up_to(len(dna) - 1) + f(len(dna)))
) * 100
return complexity_score
def get_sequence_similarity(
original: str, predicted: str, truncate: bool = True, window_length: int = 1
) -> float:
"""
Calculate the sequence similarity between two sequences.
Args:
original (str): The original sequence.
predicted (str): The predicted sequence.
truncate (bool): If True, truncate the original sequence to match the length
of the predicted sequence.
window_length (int): Length of the window for comparison (1 for amino acids,
3 for codons).
Returns:
float: The sequence similarity as a percentage.
Preconditions:
len(predicted) <= len(original).
"""
if not truncate and len(original) != len(predicted):
raise ValueError(
"Set truncate to True if the length of sequences do not match."
)
identity = 0.0
original = original.strip()
predicted = predicted.strip()
if truncate:
original = original[: len(predicted)]
if window_length == 1:
# Simple comparison for amino acid
for i in range(len(predicted)):
if original[i] == predicted[i]:
identity += 1
else:
# Comparison for substrings based on window_length
for i in range(0, len(original) - window_length + 1, window_length):
if original[i : i + window_length] == predicted[i : i + window_length]:
identity += 1
return (identity / (len(predicted) / window_length)) * 100
def scan_for_restriction_sites(seq: str, sites: List[str] = ['GAATTC', 'GGATCC', 'AAGCTT']) -> int:
"""
Scans for a list of restriction enzyme sites in a DNA sequence.
"""
return sum(seq.upper().count(site.upper()) for site in sites)
def count_negative_cis_elements(seq: str, motifs: List[str] = ['TATAAT', 'TTGACA', 'AGCTAGT']) -> int:
"""
Counts occurrences of negative cis-regulatory elements in a DNA sequence.
"""
return sum(seq.upper().count(m.upper()) for m in motifs)
def calculate_homopolymer_runs(seq: str, max_len: int = 8) -> int:
"""
Calculates the number of homopolymer runs longer than a given length.
"""
import re
min_len = max_len + 1
return len(re.findall(r'(A{%d,}|T{%d,}|G{%d,}|C{%d,})' % (min_len, min_len, min_len, min_len), seq.upper()))
def get_min_max_profile(
dna: str,
codon_frequencies: Dict[str, Tuple[List[str], List[float]]],
window_size: int = 18,
) -> List[float]:
"""
Calculate the %MinMax profile for a DNA sequence. This is a list of
%MinMax values for sliding windows across the sequence.
Args:
dna (str): The DNA sequence.
codon_frequencies (Dict[str, Tuple[List[str], List[float]]]): Codon
frequency distribution per amino acid.
window_size (int): Size of the window to calculate %MinMax.
Returns:
List[float]: List of %MinMax values for the sequence.
"""
return get_min_max_percentage(dna, codon_frequencies, window_size)
def calculate_dtw_distance(profile1: List[float], profile2: List[float]) -> float:
"""
Calculates the Dynamic Time Warping (DTW) distance between two profiles.
Args:
profile1 (List[float]): The first profile (e.g., %MinMax of generated sequence).
profile2 (List[float]): The second profile (e.g., %MinMax of natural sequence).
Returns:
float: The DTW distance between the two profiles.
"""
from dtw import dtw
import numpy as np
# Ensure profiles are numpy arrays and handle potential None and NaN values
p1 = np.array([v for v in profile1 if v is not None and not np.isnan(v)]).reshape(
-1, 1
)
p2 = np.array([v for v in profile2 if v is not None and not np.isnan(v)]).reshape(
-1, 1
)
if len(p1) == 0 or len(p2) == 0:
return np.inf # Return infinity if one of the profiles is empty
alignment = dtw(p1, p2, keep_internals=True)
return alignment.distance # type: ignore
def get_ecoli_tai_weights():
"""
Returns a dictionary of tAI weights for E. coli based on tRNA gene copy numbers.
These weights are pre-calculated based on the relative adaptiveness of each codon.
"""
codons = [
"TTT", "TTC", "TTA", "TTG", "TCT", "TCC", "TCA", "TCG", "TAT", "TAC",
"TGT", "TGC", "TGG", "CTT", "CTC", "CTA", "CTG", "CCT", "CCC", "CCA",
"CCG", "CAT", "CAC", "CAA", "CAG", "CGT", "CGC", "CGA", "CGG", "ATT",
"ATC", "ATA", "ACT", "ACC", "ACA", "ACG", "AAT", "AAC", "AAA", "AAG",
"AGT", "AGC", "AGA", "AGG", "GTT", "GTC", "GTA", "GTG", "GCT", "GCC",
"GCA", "GCG", "GAT", "GAC", "GAA", "GAG", "GGT", "GGC", "GGA", "GGG"
]
weights = [
0.1966667, 0.3333333, 0.1666667, 0.2200000, 0.1966667, 0.3333333,
0.1666667, 0.2200000, 0.2950000, 0.5000000, 0.09833333, 0.1666667,
0.2200000, 0.09833333, 0.1666667, 0.1666667, 0.7200000, 0.09833333,
0.1666667, 0.1666667, 0.2200000, 0.09833333, 0.1666667, 0.3333333,
0.4400000, 0.6666667, 0.4800000, 0.00006666667, 0.1666667, 0.2950000,
0.5000000, 0.01833333, 0.1966667, 0.3333333, 0.1666667, 0.3866667,
0.3933333, 0.6666667, 1.0000000, 0.3200000, 0.09833333, 0.1666667,
0.1666667, 0.2200000, 0.1966667, 0.3333333, 0.8333333, 0.2666667,
0.1966667, 0.3333333, 0.5000000, 0.1600000, 0.2950000, 0.5000000,
0.6666667, 0.2133333, 0.3933333, 0.6666667, 0.1666667, 0.2200000
]
return dict(zip(codons, weights))
def calculate_tAI(sequence: str, tai_weights: Dict[str, float]) -> float:
"""
Calculates the tRNA Adaptation Index (tAI) for a given DNA sequence.
Args:
sequence (str): The DNA sequence to analyze.
tai_weights (Dict[str, float]): A dictionary of tAI weights for each codon.
Returns:
float: The tAI value for the sequence.
"""
from scipy.stats.mstats import gmean
codons = [sequence[i:i+3] for i in range(0, len(sequence), 3)]
# Filter out stop codons and codons not in weights
weights = [tai_weights[codon] for codon in codons if codon in tai_weights and tai_weights[codon] > 0]
if not weights:
return 0.0
return gmean(weights)
def calculate_ENC(sequence: str) -> float:
"""
Calculate the Effective Number of Codons (ENC) for a DNA sequence.
Uses the codonbias library implementation based on Wright (1990).
Args:
sequence (str): The DNA sequence.
Returns:
float: The ENC value for the sequence.
"""
try:
from codonbias.scores import EffectiveNumberOfCodons
# Initialize ENC calculator
enc_calculator = EffectiveNumberOfCodons(
k_mer=1, # Standard codon analysis
bg_correction=True, # Use background correction
robust=True, # Use robust calculation
genetic_code=1 # Standard genetic code
)
# Calculate ENC for the sequence
enc_value = enc_calculator.get_score(sequence)
return float(enc_value)
except ImportError:
raise ImportError("codonbias library is required for ENC calculation. Install with: pip install codonbias")
except Exception as e:
# Fallback to a simple ENC approximation if library fails
print(f"Warning: ENC calculation failed with error: {e}. Using approximation.")
return 45.0 # Typical E. coli ENC value as fallback
def calculate_CPB(sequence: str, reference_sequences: Optional[List[str]] = None) -> float:
"""
Calculate the Codon Pair Bias (CPB) for a DNA sequence.
Uses the codonbias library implementation based on Coleman et al. (2008).
Args:
sequence (str): The DNA sequence.
reference_sequences (List[str]): Reference sequences for calculating expected values.
If None, uses a default E. coli reference.
Returns:
float: The CPB value for the sequence.
"""
try:
from codonbias.scores import CodonPairBias
# Use provided reference sequences or default
if reference_sequences is None:
# Use the input sequence as reference if none provided
reference_sequences = [sequence]
# Initialize CPB calculator with reference sequences
cpb_calculator = CodonPairBias(
ref_seq=reference_sequences,
k_mer=2, # Codon pairs
genetic_code=1, # Standard genetic code
ignore_stop=True, # Ignore stop codons
pseudocount=1 # Pseudocount for unseen pairs
)
# Calculate CPB for the sequence
cpb_value = cpb_calculator.get_score(sequence)
return float(cpb_value)
except ImportError:
raise ImportError("codonbias library is required for CPB calculation. Install with: pip install codonbias")
except Exception as e:
# Fallback calculation if library fails
print(f"Warning: CPB calculation failed with error: {e}. Using approximation.")
return 0.0 # Neutral CPB as fallback
def calculate_SCUO(sequence: str) -> float:
"""
Calculate the Synonymous Codon Usage Order (SCUO) for a DNA sequence.
Uses the GCUA library implementation based on information theory.
Args:
sequence (str): The DNA sequence.
Returns:
float: The SCUO value (0-1, where 1 indicates maximum bias).
"""
# Self-contained SCUO implementation (no external GCUA dependency).
# Based on Wan et al., 2004 information-theoretic definition.
from math import log2 # local import to avoid global cost
try:
# Build standard genetic code mapping using built-in tables (Biopython optional).
# Fall back to hard-coded table if Biopython absent.
try:
from Bio.Data import CodonTable # type: ignore
codon_to_aa = CodonTable.unambiguous_dna_by_id[1].forward_table
except Exception:
codon_to_aa = {
# Partial table sufficient for SCUO calculation; stop codons omitted.
'TTT': 'F', 'TTC': 'F', 'TTA': 'L', 'TTG': 'L',
'CTT': 'L', 'CTC': 'L', 'CTA': 'L', 'CTG': 'L',
'ATT': 'I', 'ATC': 'I', 'ATA': 'I', 'ATG': 'M',
'GTT': 'V', 'GTC': 'V', 'GTA': 'V', 'GTG': 'V',
'TCT': 'S', 'TCC': 'S', 'TCA': 'S', 'TCG': 'S',
'CCT': 'P', 'CCC': 'P', 'CCA': 'P', 'CCG': 'P',
'ACT': 'T', 'ACC': 'T', 'ACA': 'T', 'ACG': 'T',
'GCT': 'A', 'GCC': 'A', 'GCA': 'A', 'GCG': 'A',
'TAT': 'Y', 'TAC': 'Y', 'TAA': '*', 'TAG': '*',
'CAT': 'H', 'CAC': 'H', 'CAA': 'Q', 'CAG': 'Q',
'AAT': 'N', 'AAC': 'N', 'AAA': 'K', 'AAG': 'K',
'GAT': 'D', 'GAC': 'D', 'GAA': 'E', 'GAG': 'E',
'TGT': 'C', 'TGC': 'C', 'TGA': '*', 'TGG': 'W',
'CGT': 'R', 'CGC': 'R', 'CGA': 'R', 'CGG': 'R',
'AGT': 'S', 'AGC': 'S', 'AGA': 'R', 'AGG': 'R',
'GGT': 'G', 'GGC': 'G', 'GGA': 'G', 'GGG': 'G',
}
# Group codons by amino acid (exclude stops)
aa_to_codons = {}
for codon, aa in codon_to_aa.items():
aa_to_codons.setdefault(aa, []).append(codon)
# Count codon occurrences in input sequence
seq = sequence.upper().replace('U', 'T')
codon_counts = {}
for i in range(0, len(seq) - len(seq) % 3, 3):
codon = seq[i:i+3]
if codon in codon_to_aa:
codon_counts[codon] = codon_counts.get(codon, 0) + 1
total_codons = sum(codon_counts.values())
if total_codons == 0:
return 0.0
scuo_sum = 0.0
for aa, codons in aa_to_codons.items():
n_codons = len(codons)
if n_codons == 1:
continue # SCUO undefined for Met/Trp
counts = [codon_counts.get(c, 0) for c in codons]
total_aa = sum(counts)
if total_aa == 0:
continue
probs = [c / total_aa for c in counts if c]
H_obs = -sum(p * log2(p) for p in probs)
H_max = log2(n_codons)
O_i = (H_max - H_obs) / H_max if H_max else 0.0
F_i = total_aa / total_codons
scuo_sum += F_i * O_i
return scuo_sum
except Exception as exc:
print(f"Warning: internal SCUO computation failed ({exc}). Returning 0.5.")
return 0.5