Ranking Biomarkers Through Mutual Information

TL;DR

This paper introduces an information theoretic framework for ranking predictive and prognostic biomarkers, proposing efficient estimators and visualization tools to improve biomarker discovery in clinical trials.

Contribution

It formalizes biomarker ranking as an optimization of mutual information, with novel estimators and visualization methods for better differentiation of biomarker types.

Findings

Proposes a formal information theoretic approach to biomarker ranking.

Develops efficient low-dimensional mutual information estimators.

Introduces a visualization tool for biomarker strength analysis.

Abstract

We study information theoretic methods for ranking biomarkers. In clinical trials there are two, closely related, types of biomarkers: predictive and prognostic, and disentangling them is a key challenge. Our first step is to phrase biomarker ranking in terms of optimizing an information theoretic quantity. This formalization of the problem will enable us to derive rankings of predictive/prognostic biomarkers, by estimating different, high dimensional, conditional mutual information terms. To estimate these terms, we suggest efficient low dimensional approximations, and we derive an empirical Bayes estimator, which is suitable for small or sparse datasets. Finally, we introduce a new visualisation tool that captures the prognostic and the predictive strength of a set of biomarkers. We believe this representation will prove to be a powerful tool in biomarker discovery.