Calculator for Mutual Information Between a Discrete and a Continuous Data Set

Abstract

Mutual information (MI) is in many ways an ideal statistic for detecting relationships between two data sets. MI is easy to calculate when both data sets are discrete, but not when one or both data sets are real-valued. An accurate method for calculating MI between two real-valued data sets was previously developed (Kraskov et al. 2004). We present an accurate method for calculating MI between one discrete data set and one real-valued data set. For example, this calculator can quantify the correlation between base methylation (a discrete variable) and gene expression level (real-valued), or the effect of a clinical procedure (boolean; discrete) on patient survival time (real-valued). We use our calculator in the context of nanopore sequencing where DNA is drawn through a MspA pore protein.Kraskov, A., Stogbauer, H., & Grassberger, P. (2004). Estimating mutual information. Physical Review E, 69(6), 066138.