8: Sequence Motifs, Position Weight Matrices & Information Content

Chapter 8: Sequence Motifs, Position Weight Matrices & Information Content

Gene regulation relies on sequence-specific DNA-binding proteins, such as transcription factors, recognizing short (6–20 bp) degenerate regulatory elements in promoters and enhancers. Because DNA-binding proteins tolerate nucleotide substitutions at non-critical positions, representing these motifs with simple consensus strings loses vital biological affinity data. Position Weight Matrices (PWMs) provide the rigorous mathematical standard for modeling sequence motifs.

1. Constructing a Position Weight Matrix (PWM)

# Given a set of aligned TATA-box binding sites:
# Step 1: Position Frequency Matrix (PFM) - Count observed bases per position
# Step 2: Position Probability Matrix (PPM) - Add pseudocounts and normalize:
#         p(b, l) = (Count(b, l) + s(b)) / (Total + sum(s))
# Step 3: Position Weight Matrix (PWM) - Log-odds ratio against background:
#         W(b, l) = log2( p(b, l) / q(b) ) where q(b) = 0.25

2. Programmatic Motif Scanning with BioPython

from Bio import motifs
from Bio.Seq import Seq

# Define a collection of binding site instances
instances = [
    Seq("TATAAA"),
    Seq("TATATA"),
    Seq("TATAAA"),
    Seq("TATAAT"),
    Seq("CATAAA"),
]

# Create motif object and build PWM
m = motifs.create(instances)
print("Position Frequency Matrix (PFM):\n", m.counts)

# Convert to Position Specific Scoring Matrix (PSSM / PWM)
pwm = m.counts.normalize(pseudocounts=0.5)
pssm = pwm.log_odds()
print("\nPSSM (Log-odds scores):\n", pssm)

# Scan a promoter sequence for candidate binding sites
promoter = Seq("CGGATATAAACCGGCTATATAATGCC")
for position, score in pssm.search(promoter, threshold=3.0):
    print(f"Discovered hit at position {position:2d} | Score = {score:.2f} | Sequence = {promoter[position:position+6]}")