kchu25 / MotifPvalue.jl

Threshold and p-value computations for Position Weight Matrices

MotifPvalue

Introduction

This package provides two useful subroutines:

• Given a score threshold α, score2pvalue calculates the p-value, i.e. the probability of a set of strings generated from a specfied background model such that each string in this set can attain a score higher than or equal to α with respect to the input pwm.
• Given a p-value pval, pval2score calculates the highest score α such that score2pvalue(pwm,α) is larger than or equal to pval.

By default, the background model is specified as i.i.d discrete uniform.

This is an implementation of:

Efficient and accurate P-value computation for Position Weight Matrices by Touzet et al. https://almob.biomedcentral.com/articles/10.1186/1748-7188-2-15

Basic examples

using MotifPvalue

# An example PWM
pwm = [-2.86995   1.3814    1.36906  -4.88479   -4.19164  -4.88479   1.36607  -5.57793   -0.336375    1.38238;
        1.36707  -5.57793  -2.86995  -0.294869  -4.88479   1.34492  -3.49849   1.26497   -2.1768     -4.88479;
       -5.57793  -4.47899  -5.57793   1.11923   -3.78584  -5.57793  -4.88479  -5.57793    0.798703   -5.57793;
       -4.19164  -5.57793  -4.88479  -1.68576    1.375    -1.88885  -3.17976  -0.798616  -0.0526305  -5.57793];

# Compute pvalue with pwm and score = 4
score2pvalue(pwm, 4)
> 0.000213623046875

# Compute the score-threshold for p-value 1e-4
pvalue2score(pwm, 1e-4)
> 5.57