Content
Summary
Motif finding identifies recurring biological patterns using consensus sequences and Position Specific Scoring Matrices (PSSMs). The process requires filtering low-complexity sequences and quantifying motif strength through Shannon entropy and information content relative to a background model.
Intuition
Finding a motif is like searching for a specific signal in noisy data. First, we define the signal using consensus sequences with IUPAC codes, which allow for biological variation at each position. Before searching, we must remove 'noise'—low-complexity repeats like poly-A tails that can cause false positives. We then use a PSSM to represent the motif probabilistically. The power of this motif is measured by how much it reduces uncertainty (entropy) compared to a random background model. High information content means the motif is highly specific and unlikely to occur by chance.
Key equations
Shannon entropy at position i, measuring uncertainty based on nucleotide probabilities.
Information content at position i, defined as the reduction in entropy relative to the maximum background entropy.
Kullback-Leibler divergence, measuring the disparity between the motif distribution and background frequencies.
Derivation
- Define motif finding as identifying recurring patterns in biological sequences.
- Represent motifs using consensus sequences with IUPAC ambiguity codes to handle variations.
- Filter low-complexity sequences using tools like DUST to prevent confounding results.
- Construct a Position Specific Scoring Matrix (PSSM) where P_ij represents the probability of nucleotide j at position i.
- Calculate Shannon entropy (H_i) to quantify uncertainty at each position.
- Determine Information Content (IC_i) as the difference between maximum background entropy and position-specific entropy.
- Use Kullback-Leibler divergence to measure the information gain relative to a background model.
Worked example
Given a motif position where A=0.8, C=0.1, G=0.1, T=0.0, and a background model where all nucleotides are equal (0.25 each), calculate the information content (IC) in bits.
- Calculate background maximum entropy: H_max = -sum(0.25 * log2(0.25)) for 4 nucleotides = 2 bits.
- Calculate position entropy H_i: H_i = -(0.8*log2(0.8) + 0.1*log2(0.1) + 0.1*log2(0.1) + 0.0*log2(0.0)). Note: 0*log2(0) is treated as 0.
- H_i ≈ -(0.8*(-0.322) + 0.1*(-3.322) + 0.1*(-3.322)) ≈ -(-0.2576 - 0.3322 - 0.3322) ≈ 0.922 bits.
- Calculate IC_i: IC_i = H_max - H_i = 2 - 0.922 = 1.078 bits.
Answer The information content is approximately 1.078 bits.
Common mistakes
- Ignoring low-complexity sequences, which can lead to false positive motif matches.
- Confusing the background model with the motif model; the background represents random genomic frequency, not the motif.
- Failing to align sequences precisely before constructing the count matrix for the PSSM.
- Using natural logarithms when bits are required, leading to incorrect units (nits vs bits).