SeqLogo

What is this?

SeqLogo is a package for plotting position weight matrices (PWMs), which is commonly used to characterize and visualize motifs — the binding sites where proteins interact with DNA or RNA.

Table of contents

• Usage
  • Plot your typical PWMs
    • Save the PWMs
  • Plot your PWMs with crosslinking tendencies
    • Multiplexed crosslinking tendencies
• Some-definitions

Usage

Plot your typical PWMs

using SeqLogo

# Given a position frequency matrix (PFM), where each column sums to 1

pfm =  [0.02  1.0  0.98  0.0   0.0   0.0   0.98  0.0   0.18  1.0
        0.98  0.0  0.02  0.19  0.0   0.96  0.01  0.89  0.03  0.0
        0.0   0.0  0.0   0.77  0.01  0.0   0.0   0.0   0.56  0.0
        0.0   0.0  0.0   0.04  0.99  0.04  0.01  0.11  0.23  0.0]

# Define the background probabilities for (A, C, G, T)

background = [0.25, 0.25, 0.25, 0.25]

logoplot(pfm, background)

will give

The function logoplot(pfm, background) produces a plot where:

• The x-axis shows the positions in the PWM.
• The y-axis shows the information content (bits) for each position.

The background is an array representing the background probabilities for A, C, G, and T. These should sum to 1. In this example, a uniform background of [0.25, 0.25, 0.25, 0.25] is used, assuming equal probabilities for each base.

You can also call:

logoplot(pfm)

to get the same results as above, where the background is set to [0.25, 0.25, 0.25, 0.25] by default.

Save the PWMs

To save your plot, use save_logoplot(pfm, background, save_name). For example:

save_logoplot(pfm, background, "tmp/logo.png")

Or simply:

save_logoplot(pfm, "tmp/logo.png")

where a uniform background of [0.25, 0.25, 0.25, 0.25] is used implicitly.

Plot your PWMs with crosslinking tendencies

Cross-linked PWMs not only display the PWM but also account for crosslinking tendencies, which are particularly relevant for RNA-binding proteins (RBPs) from CLIP-Seq.

To plot these, you'll need to estimate the crosslinking tendencies along with the PFM. For a PFM with $L$ columns, provide a $K \times L$ matrix $C$, where $\sum_{k,\ell}C_{k\ell} \leq 1$.

For example, when $K=1$:

C = [0.01  0.04  0.05  0.0  0.74  0.05  0.03  0.05  0.03  0.0] 

and the background:

background = [0.25, 0.25, 0.25, 0.25]

You can then plot the cross-linked PWM using:

logoplotwithcrosslink(pfm, background, C; rna=true)

This will generate:

Setting the tag rna=true will change the logo from using thymine T to uracil U.

Alternatively, you can use:

logoplotwithcrosslink(pfm, C; rna=true)

which will automatically assume a uniform background of [0.25, 0.25, 0.25, 0.25].

Multiplexed crosslinking tendencies

Multiplexed crosslinking tendencies occur when multiple crosslinking signatures are present in the dataset. Each signature can be applied to each sequence before performing motif discovery tasks. This situation corresponds to cases where the crosslink matrix $C$ has more than one row, i.e., $K > 1$."

Suppose we have

C2 = [0.01  0.02  0.03  0.0   0.37  0.03  0.02  0.03  0.02  0.0
      0.03  0.0   0.11  0.04  0.26  0.0   0.03  0.01  0.02  0.02]

Now, using

logoplotwithcrosslink(pfm, background, C2; rna=true)

You'd get

Here, different colors indicate different crosslinking signatures, and their height is proportional to the crosslinking tendency at each position in the PWM.

Some-definitions

Definition of Information Content in PWMs

In a position weight matrix (PWM), the "letter height", or more formally, the information content $IC(\cdot)$ of the $i$-th column $c_i$, quantifies how conserved the nucleotides are at that position compared to a background model. It is calculated using the formula:

$$ IC(c_i) = \sum_{\alpha}f_{\alpha i}\log_2 \frac{f_{\alpha i}}{\beta_\alpha} $$

where

• $f_{\alpha i}$ the frequency of nucleotide $\alpha\in\Set{A,C,G,T}$ at the $i$-th column of a PWM.
• $\beta_\alpha$ denotes the genomic background frequency of nucleotide $\alpha$.

Default genomic background

By default, the background model assumes a uniform distribution of nucleotides, with each nucleotide having a frequency of $\beta=(0.25, 0.25,0.25,0.25)$. In this case, the information content $IC(c_i)$ simplifies to:

$$IC(c_i)=2+\sum_{\alpha}f_{\alpha i}\log_2 f_{\alpha i}$$

This formula illustrates why the y-axis of the plot ranges from $0$ to $2$.

Acknowledgement

This code repo modifies the code using the work from https://github.com/BenjaminDoran/LogoPlots.jl.