Reconstructing gene expression and knockout effect scores from DNA mutation (Mut2Ex): methodology and application to cancer prediction problems

TL;DR

This paper introduces Mut2Ex, a method that reconstructs gene expression and knockout effect scores from limited mutational data to improve clinical outcome predictions in cancer, demonstrating significant accuracy gains.

Contribution

The paper presents a novel approach to augment mutational data with reconstructed gene expression and effect scores, enhancing predictive modeling in cancer research.

Findings

Significant improvement in predictive accuracy over raw mutational models

Reconstructed scores yield results comparable to real expression or effect profiles

Method effectively leverages multi-modal relationships for better predictions

Abstract

Building prediction models for outcomes of clinical relevance when only a limited number of mutational features are available causes considerable challenges due to the sparseness and low-dimensionality of the data. In this article, we present a method to augment the predictive power of these features by leveraging multi-modal associative relationships between an individual's mutational profile and their corresponding gene expression or knockout effect profiles. We can thus reconstruct expression or effect scores for genes of interest from the available mutation features and then use this reconstructed representation directly to model and predict clinical outcomes. We show that our method produces significant improvements in predictive accuracy compared to models utilizing only the raw mutational data, and results in conclusions comparable to those obtained using real expression or effect profiles.